rfc9522xml2.original.xml   rfc9522.xml 
<?xml version="1.0" encoding="utf-8"?> <?xml version="1.0" encoding="UTF-8"?>
<?xml-stylesheet type="text/xsl" href="rfc2629.xslt" ?>
<?rfc toc="yes"?>
<?rfc sortrefs="yes"?>
<?rfc symrefs="yes"?>
<?rfc comments="yes"?>
<!DOCTYPE rfc SYSTEM "rfc2629.dtd" [
<!ENTITY RFC0791 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.0791.xml">
<!ENTITY RFC1102 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.1102.xml">
<!ENTITY RFC1104 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.1104.xml">
<!ENTITY RFC2205 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2205.xml">
<!ENTITY RFC2330 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2330.xml">
<!ENTITY RFC2386 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2386.xml">
<!ENTITY RFC2474 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2474.xml">
<!ENTITY RFC2475 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2475.xml">
<!ENTITY RFC2597 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2597.xml">
<!ENTITY RFC2678 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2678.xml">
<!ENTITY RFC2702 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2702.xml">
<!ENTITY RFC2722 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2722.xml">
<!ENTITY RFC2753 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2753.xml">
<!ENTITY RFC2961 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2961.xml">
<!ENTITY RFC2998 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.2998.xml">
<!ENTITY RFC3031 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3031.xml">
<!ENTITY RFC3086 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3086.xml">
<!ENTITY RFC3124 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3124.xml">
<!ENTITY RFC3168 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3168.xml">
<!ENTITY RFC3175 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3175.xml">
<!ENTITY RFC3198 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3198.xml">
<!ENTITY RFC3209 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3209.xml">
<!ENTITY RFC3270 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3270.xml">
<!ENTITY RFC3272 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3272.xml">
<!ENTITY RFC3469 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3469.xml">
<!ENTITY RFC3473 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3473.xml">
<!ENTITY RFC3630 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3630.xml">
<!ENTITY RFC3945 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.3945.xml">
<!ENTITY RFC4090 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4090.xml">
<!ENTITY RFC4124 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4124.xml">
<!ENTITY RFC4203 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4203.xml">
<!ENTITY RFC4271 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4271.xml">
<!ENTITY RFC4340 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4340.xml">
<!ENTITY RFC4461 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4461.xml">
<!ENTITY RFC4594 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4594.xml">
<!ENTITY RFC4655 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4655.xml">
<!ENTITY RFC4872 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4872.xml">
<!ENTITY RFC4873 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4873.xml">
<!ENTITY RFC4875 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.4875.xml">
<!ENTITY RFC5151 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5151.xml">
<!ENTITY RFC5250 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5250.xml">
<!ENTITY RFC5305 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5305.xml">
<!ENTITY RFC5329 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5329.xml">
<!ENTITY RFC5331 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5331.xml">
<!ENTITY RFC5357 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5357.xml">
<!ENTITY RFC5394 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5394.xml">
<!ENTITY RFC5440 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5440.xml">
<!ENTITY RFC5470 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5470.xml">
<!ENTITY RFC5472 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5472.xml">
<!ENTITY RFC5541 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5541.xml">
<!ENTITY RFC5557 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5557.xml">
<!ENTITY RFC5559 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5559.xml">
<!ENTITY RFC5623 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5623.xml">
<!ENTITY RFC5664 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5664.xml">
<!ENTITY RFC5671 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5671.xml">
<!ENTITY RFC5693 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.5693.xml">
<!ENTITY RFC6107 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.6107.xml">
<!ENTITY RFC6119 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.6119.xml">
<!ENTITY RFC6241 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.6241.xml">
<!ENTITY RFC6372 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.6372.xml">
<!ENTITY RFC6374 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.6374.xml">
<!ENTITY RFC6601 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.6601.xml">
<!ENTITY RFC6805 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.6805.xml">
<!ENTITY RFC7011 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7011.xml">
<!ENTITY RFC7149 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7149.xml">
<!ENTITY RFC7285 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7285.xml">
<!ENTITY RFC7390 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7390.xml">
<!ENTITY RFC7426 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7426.xml">
<!ENTITY RFC7471 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7471.xml">
<!ENTITY RFC7491 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7491.xml">
<!ENTITY RFC7551 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7551.xml">
<!ENTITY RFC7567 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7567.xml">
<!ENTITY RFC7679 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7679.xml">
<!ENTITY RFC7680 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7680.xml">
<!ENTITY RFC7665 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7665.xml">
<!ENTITY RFC7752 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7752.xml">
<!ENTITY RFC7923 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7923.xml">
<!ENTITY RFC7926 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7926.xml">
<!ENTITY RFC7950 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.7950.xml">
<!ENTITY RFC8033 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8033.xml">
<!ENTITY RFC8034 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8034.xml">
<!ENTITY RFC8040 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8040.xml">
<!ENTITY RFC8051 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8051.xml">
<!ENTITY RFC8189 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8189.xml">
<!ENTITY RFC8231 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8231.xml">
<!ENTITY RFC8259 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8259.xml">
<!ENTITY RFC8279 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8279.xml">
<!ENTITY RFC8281 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8281.xml">
<!ENTITY RFC8283 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8283.xml">
<!ENTITY RFC8290 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8290.xml">
<!ENTITY RFC8402 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8402.xml">
<!ENTITY RFC8453 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8453.xml">
<!ENTITY RFC8570 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8570.xml">
<!ENTITY RFC8571 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8571.xml">
<!ENTITY RFC8655 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8655.xml">
<!ENTITY RFC8664 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8664.xml">
<!ENTITY RFC8684 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8684.xml">
<!ENTITY RFC8685 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8685.xml">
<!ENTITY RFC8795 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8795.xml">
<!ENTITY RFC8803 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8803.xml">
<!ENTITY RFC8896 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8896.xml">
<!ENTITY RFC8938 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8938.xml">
<!ENTITY RFC8955 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8955.xml">
<!ENTITY RFC8972 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.8972.xml">
<!ENTITY RFC9000 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9000.xml">
<!ENTITY RFC9023 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9023.xml">
<!ENTITY RFC9040 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9040.xml">
<!ENTITY RFC9113 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9113.xml">
<!ENTITY RFC9256 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9256.xml">
<!ENTITY RFC9262 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9262.xml">
<!ENTITY RFC9298 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9298.xml">
<!ENTITY RFC9315 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9315.xml">
<!ENTITY RFC9332 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9332.xml">
<!ENTITY RFC9350 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9350.xml">
<!ENTITY RFC9439 SYSTEM "http://xml.resource.org/public/rfc/bibxml/reference.RFC
.9439.xml">
<!ENTITY I-D.ietf-bess-evpn-unequal-lb SYSTEM "http://xml2rfc.ietf.org/public/rf
c/bibxml3/reference.I-D.ietf-bess-evpn-unequal-lb">
<!ENTITY I-D.ietf-idr-performance-routing SYSTEM "http://xml2rfc.ietf.org/public
/rfc/bibxml3/reference.I-D.ietf-idr-performance-routing">
<!ENTITY I-D.ietf-idr-segment-routing-te-policy SYSTEM "http://xml2rfc.ietf.org/
public/rfc/bibxml3/reference.I-D.ietf-idr-segment-routing-te-policy">
<!ENTITY I-D.ietf-lsr-ip-flexalgo SYSTEM "http://xml2rfc.ietf.org/public/rfc/bib
xml3/reference.I-D.ietf-lsr-ip-flexalgo">
<!ENTITY I-D.ietf-quic-multipath SYSTEM "http://xml2rfc.ietf.org/public/rfc/bibx
ml3/reference.I-D.ietf-quic-multipath">
<!ENTITY I-D.ietf-rtgwg-segment-routing-ti-lfa SYSTEM "http://xml2rfc.ietf.org/
public/rfc/bibxml3/reference.I-D.ietf-rtgwg-segment-routing-ti-lfa">
<!ENTITY I-D.ietf-teas-enhanced-vpn SYSTEM "http://xml2rfc.ietf.org/public/rfc/b
ibxml3/reference.I-D.ietf-teas-enhanced-vpn">
<!ENTITY I-D.ietf-tewg-qos-routing SYSTEM "http://xml2rfc.ietf.org/public/rfc/bi
bxml3/reference.I-D.ietf-tewg-qos-routing">
<!ENTITY I-D.ietf-tsvwg-multipath-dccp SYSTEM "http://xml2rfc.ietf.org/public/rf
c/bibxml3/reference.I-D.ietf-tsvwg-multipath-dccp">
<!ENTITY I-D.ietf-teas-ietf-network-slices SYSTEM "http://xml2rfc.ietf.org/publi
c/rfc/bibxml3/reference.I-D.ietf-teas-ietf-network-slices">
<!DOCTYPE rfc [
<!ENTITY nbsp "&#160;">
<!ENTITY zwsp "&#8203;">
<!ENTITY nbhy "&#8209;">
<!ENTITY wj "&#8288;">
]> ]>
<rfc docName="draft-ietf-teas-rfc3272bis-27" category="info" obsoletes="3272" ip <rfc xmlns:xi="http://www.w3.org/2001/XInclude" docName="draft-ietf-teas-rfc3272
r="trust200902"> bis-27" number="9522" submissionType="IETF" category="info" consensus="true" obs
oletes="3272" ipr="trust200902" updates="" xml:lang="en" tocInclude="true" sortR
<front> efs="true" symRefs="true" version="3">
<title abbrev="Overview and Principles of Internet TE">Overview and Principles o
f Internet Traffic Engineering</title>
<author initials="A." surname="Farrel" fullname="Adrian Farrel" role="editor">
<organization>Old Dog Consulting</organization>
<address>
<email>adrian@olddog.co.uk</email>
</address>
</author>
<date year="2023" />
<workgroup>TEAS Working Group</workgroup>
<abstract>
<t>This document describes the principles of traffic engineering (TE) in the
Internet. The document is intended to promote better understanding
of the issues surrounding traffic engineering in IP networks and the networ
ks
that support IP networking, and to provide a common basis for the developme
nt
of traffic engineering capabilities for the Internet. The principles,
architectures, and methodologies for performance evaluation and performance
optimization of operational networks are also discussed.</t>
<t>This work was first published as RFC 3272 in May 2002. This document
obsoletes RFC 3272 by making a complete update to bring the text in line
with best current practices for Internet traffic engineering and to
include references to the latest relevant work in the IETF.</t>
</abstract>
</front>
<middle>
<section anchor="INTRO" title="Introduction">
<t>This document describes the principles of Internet traffic engineering (TE) <front>
. <title abbrev="Overview and Principles of Internet TE">Overview and Principl
The objective of the document is to articulate the general issues and es of Internet Traffic Engineering</title>
principles for Internet TE, and where appropriate to <seriesInfo name="RFC" value="9522"/>
provide recommendations, guidelines, and options for the development <author initials="A." surname="Farrel" fullname="Adrian Farrel" role="editor
of preplanned (offline) and dynamic (online) Internet TE ">
capabilities and support systems.</t> <organization>Old Dog Consulting</organization>
<address>
<email>adrian@olddog.co.uk</email>
</address>
</author>
<date year="2024" month="January"/>
<area>rtg</area>
<workgroup>teas</workgroup>
<t>Even though Internet TE is most effective when <keyword>Policy</keyword>
applied end-to-end, the focus of this document is TE <keyword>Path steering</keyword>
within a given domain (such as an autonomous system). However, because <keyword>Resource management</keyword>
a preponderance of Internet traffic tends to originate in one autonomous <keyword>Network engineering</keyword>
system and terminate in another, this document also provides an overview of <keyword>Network performance optimization</keyword>
aspects pertaining to inter-domain TE.</t>
<t>This document provides a terminology and taxonomy for describing and <abstract>
<t>This document describes the principles of traffic engineering (TE) in
the Internet. The document is intended to promote better understanding
of the issues surrounding traffic engineering in IP networks and the
networks that support IP networking and to provide a common basis for
the development of traffic-engineering capabilities for the Internet.
The principles, architectures, and methodologies for performance
evaluation and performance optimization of operational networks are also
discussed.</t>
<t>This work was first published as RFC 3272 in May 2002. This document
obsoletes RFC 3272 by making a complete update to bring the text in line
with best current practices for Internet traffic engineering and to
include references to the latest relevant work in the IETF.</t>
</abstract>
</front>
<middle>
<section anchor="INTRO" numbered="true" toc="default">
<name>Introduction</name>
<t>This document describes the principles of Internet traffic
engineering (TE). The objective of the document is to articulate the
general issues and principles for Internet TE and, where appropriate, to
provide recommendations, guidelines, and options for the development of
preplanned (offline) and dynamic (online) Internet TE capabilities and
support systems.</t>
<t>Even though Internet TE is most effective when applied end-to-end,
the focus of this document is TE within a given domain (such as an
Autonomous System (AS)). However, because a preponderance of Internet
traffic tends to originate in one AS and terminate in
another, this document also provides an overview of aspects pertaining
to inter-domain TE.</t>
<t>This document provides terminology and a taxonomy for describing and
understanding common Internet TE concepts.</t> understanding common Internet TE concepts.</t>
<t>This work was first published as <xref target="RFC3272"
<t>This work was first published as <xref target="RFC3272"/> in May 2002. format="default"/> in May 2002. This document obsoletes <xref
This document obsoletes <xref target="RFC3272"/> by making a complete target="RFC3272" format="default"/> by making a complete update to bring
update to bring the text in line with best current practices for Internet the text in line with best current practices for Internet TE and to
TE and to include references to the latest relevant work include references to the latest relevant work in the IETF. It is worth
in the IETF. It is worth noting around three fifths of the RFCs referenced noting around three-fifths of the RFCs referenced in this document
in this document post-date the publication of RFC 3272. <xref target="CHAN postdate the publication of <xref target="RFC3272" format="default"/>.
GES" /> <xref target="CHANGES" format="default"/> provides a summary of changes
provides a summary of changes between RFC 3272 and this document.</t> between <xref target="RFC3272" format="default"/> and this document.</t>
<section anchor="WHATTE" numbered="true" toc="default">
<section anchor="WHATTE" title="What is Internet Traffic Engineering?"> <name>What is Internet Traffic Engineering?</name>
<t>One of the most significant functions performed in the Internet is
<t>One of the most significant functions performed in the Internet is the the routing and forwarding of traffic from ingress nodes to egress
routing and forwarding of traffic from ingress nodes to egress nodes. Th nodes. Therefore, one of the most distinctive functions performed by
erefore, one Internet traffic engineering is the control and optimization of these
of the most distinctive functions performed by Internet traffic routing and forwarding functions, to steer traffic through the
engineering is the control and optimization of these routing and forwardi network.</t>
ng functions, to <t>Internet traffic engineering is defined as that aspect of Internet
steer traffic through the network.</t> network engineering dealing with the issues of performance evaluation
and performance optimization of operational IP networks. Traffic
<t>Internet traffic engineering is defined as that aspect of Internet engineering encompasses the application of technology and scientific
network engineering dealing with the issues of performance evaluation principles to the measurement, characterization, modeling, and control
and performance optimization of operational IP networks. Traffic of Internet traffic <xref target="RFC2702" format="default"/> <xref
engineering encompasses the application of technology and scientific target="AWD2" format="default"/>.</t>
principles to the measurement, characterization, modeling, and <t>It is the performance of the network as seen by end users of
control of Internet traffic <xref target="RFC2702"/>, <xref target="AWD2" network services that is paramount. The characteristics visible to
/>.</t> end users are the emergent properties of the network, which are the
characteristics of the network when viewed as a whole. A central goal
<t>It is the performance of the network as seen by end users of network of the service provider, therefore, is to enhance the emergent
services that is paramount. The characteristics visible to end users properties of the network while taking economic considerations into
are the emergent properties of the network, which are the characteristics account. This is accomplished by addressing traffic-oriented
of the network when viewed as a whole. A central goal of the service performance requirements while utilizing network resources without
provider, therefore, is to enhance the emergent properties of the network excessive waste and in a reliable way. Traffic-oriented performance
while taking economic considerations into account. This is accomplished measures include delay, delay variation, packet loss, and
by addressing traffic oriented performance requirements while utilizing throughput.</t>
network resources without excessive waste and in a reliable way. Traffic <t>Internet TE responds to network events (such as link or node
oriented failures, reported or predicted network congestion, planned
performance measures include delay, delay variation, packet loss, and maintenance, service degradation, planned changes in the traffic
throughput.</t> matrix, etc.). Aspects of capacity management respond at intervals
ranging from days to years. Routing control functions operate at
<t>Internet TE responds to network events (such as link or intervals ranging from milliseconds to days. Packet-level processing
node failures, reported or predicted network congestion, planned maintena functions operate at very fine levels of temporal resolution (up to
nce, milliseconds) while reacting to statistical measures of the real-time
service degradation, planned changes in the traffic matrix, etc.). Aspec behavior of traffic.</t>
ts <t>Thus, the optimization aspects of TE can be viewed from a control
of capacity management respond at intervals ranging from days to years. perspective and can be both proactive and reactive. In the proactive
Routing control functions operate at intervals ranging from milliseconds case, the TE control system takes preventive action to protect against
to days. Packet level processing functions operate at very fine levels o predicted unfavorable future network states, for example, by
f engineering backup paths.
temporal resolution (up to milliseconds) while reacting to It may also take action that will lead to a
statistical measures of the real-time behavior of traffic.</t> more desirable future network state. In the reactive case, the
control system responds to correct issues and adapt to network
<t>Thus, the optimization aspects of TE can be viewed events, such as routing after failure.</t>
from a control perspective, and can be both proactive and reactive. <t>Another important objective of Internet TE is to facilitate
In the proactive case, the TE control system takes reliable network operations <xref target="RFC2702" format="default"/>.
preventive action to protect against predicted unfavorable future Reliable network operations can be facilitated by providing mechanisms
network states, for example, by engineering backup paths. It may that enhance network integrity and by embracing policies emphasizing
also take action that will lead to a more desirable future network network survivability. This reduces the vulnerability of services to
state. In the reactive case, the control system responds to correct outages arising from errors, faults, and failures occurring within the
issues and adapt to network events, such as routing after failure.</t> network infrastructure.</t>
<t>The optimization aspects of TE can be achieved
<t>Another important objective of Internet TE is to
facilitate reliable network operations <xref target="RFC2702"/>.
Reliable network operations can be facilitated by providing mechanisms
that enhance network integrity and by embracing policies emphasizing
network survivability. This reduces the vulnerability of services
to outages arising from errors, faults, and failures occurring within
the network infrastructure.</t>
<t>The optimization aspects of TE can be achieved
through capacity management and traffic management. In this through capacity management and traffic management. In this
document, capacity management includes capacity planning, routing document, capacity management includes capacity planning, routing
control, and resource management. Network resources of particular control, and resource management. Network resources of particular
interest include link bandwidth, buffer space, and computational interest include link bandwidth, buffer space, and computational
resources. In this document, traffic management includes: resources. In this document, traffic management includes:
<list style="numbers"> </t>
<t>Nodal traffic control functions such as traffic conditioning, <ol spacing="normal">
queue management, and scheduling.</t> <li>Nodal traffic control functions, such as traffic conditioning,
<t>Other functions that regulate the flow of traffic through the networ queue management, and scheduling.</li>
k <li>Other functions that regulate the flow of traffic through
or that arbitrate access to network resources between different the network or that arbitrate access to network resources between
packets or between different traffic flows.</t> different packets or between different traffic flows.</li>
</list></t> </ol>
<t>One major challenge of Internet TE is the realization of automated
<t>One major challenge of Internet TE is the control capabilities that adapt quickly and cost-effectively to
realization of automated control capabilities that adapt quickly and significant changes in network state, while still maintaining
cost effectively to significant changes in network state, while stability of the network. Performance evaluation can assess the
still maintaining stability of the network. Performance evaluation effectiveness of TE methods, and the results of this evaluation can be
can assess the effectiveness of TE methods, and the used to identify existing problems, guide network reoptimization, and
results of this evaluation can be used to identify existing problems, aid in the prediction of potential future problems. However, this
guide network re-optimization, and aid in the prediction of potential process can also be time-consuming and may not be suitable to act on
future problems. However, this process can also be time consuming and ma short-lived changes in the network.</t>
y <t>Performance evaluation can be achieved in many different ways. The
not be suitable to act on short-lived changes in the network.</t>
<t>Performance evaluation can be achieved in many different ways. The
most notable techniques include analytic methods, simulation, and most notable techniques include analytic methods, simulation, and
empirical methods based on measurements.</t> empirical methods based on measurements.</t>
<t>Traffic engineering comes in two flavors:
</t>
<ul spacing="normal">
<li>A background process that constantly monitors traffic and
network conditions and optimizes the use of resources to
improve performance.</li>
<li>A form of a pre-planned traffic distribution that is considered
optimal.</li>
</ul>
<t>Traffic engineering comes in two flavors: <t>In the latter case, any deviation from the optimum distribution
<list style="symbols"> (e.g., caused by a fiber cut) is reverted upon repair without further
<t>A background process that constantly monitors traffic and network optimization. However, this form of TE relies upon the notion that
conditions, and optimizes the use of resources to improve performanc the planned state of the network is optimal. Hence,
e.</t> there are two levels of TE in such a mode: </t>
<t>A form of a pre-planned traffic distribution that is considered opti <ul spacing="normal">
mal.</t> <li>The TE-planning task to enable optimum traffic distribution.</li>
</list> <li>The routing and forwarding tasks that keep traffic flows
In the latter case, any deviation from the optimum attached to the pre-planned distribution.</li>
distribution (e.g., caused by a fiber cut) is reverted upon repair withou </ul>
t <t>As a general rule, TE concepts and mechanisms must be sufficiently
further optimization. However, this form of TE relies specific and well-defined to address known requirements but
upon the notion that the planned state of the network is optimal. Hence, simultaneously flexible and extensible to accommodate
in such a mode there are two levels of TE: the TE-planning unforeseen future demands (see <xref target="HIGHOBJ"
task to enable optimum traffic distribution, and the routing and forwardi format="default"/>).</t>
ng </section>
tasks that keep traffic flows attached to the pre-planned distribution.</ <section anchor="COMPONENTS" numbered="true" toc="default">
t> <name>Components of Traffic Engineering</name>
<t>As mentioned in <xref target="WHATTE" format="default"/>, Internet
<t>As a general rule, TE concepts and mechanisms must be traffic engineering provides performance optimization of IP networks
sufficiently specific and well-defined to address known requirements, but while utilizing network resources economically and reliably. Such
simultaneously flexible and extensible to accommodate unforeseen future optimization is supported at the control/controller level and within
demands (see <xref target="HIGHOBJ" />).</t> the data/forwarding plane.</t>
<t>The key elements required in any TE solution are as follows:
</section> </t>
<ol spacing="normal" type="1"><li>Policy</li>
<section anchor="COMPONENTS" title="Components of Traffic Engineering"> <li>Path steering</li>
<li>Resource management</li>
<t>As mentioned in <xref target="WHATTE" />, Internet traffic engineering pr </ol>
ovides <t>Some TE solutions rely on these elements to a lesser or greater
performance optimization of IP networks while utilizing network resources extent. Debate remains about whether a solution can truly be
economically and reliably. Such optimization is supported at the called "TE" if it does not include all of these elements. For the sake
control/controller level and within the data/forwarding plane.</t> of this document, we assert that all TE solutions must include some
aspects of all of these elements. Other solutions can be classed as
<t>The key elements required in any TE solution are as follows: "partial TE" and also fall in scope of this document.</t>
<list style="numbers"> <t>Policy allows for the selection of paths (including next hops)
<t>Policy</t> based on information beyond basic reachability. Early definitions of
<t>Path steering</t> routing policy, e.g., <xref target="RFC1102" format="default"/> and
<t>Resource management</t> <xref target="RFC1104" format="default"/>, discuss routing policy
</list></t> being applied to restrict access to network resources at an aggregate
level. BGP is an example of a commonly used mechanism for applying
<t>Some TE solutions rely on these elements to a lesser or greater extent. such policies; see <xref target="RFC4271" format="default"/> and <xref
Debate remains about whether a solution can truly be called TE target="RFC8955" format="default"/>. In the TE context, policy
if it does not include all of these elements. For the sake of this docum decisions are made within the control plane or by controllers in the
ent, management plane and govern the selection of paths. Examples can be
we assert that all TE solutions must include some aspects of all of these found in <xref target="RFC4655" format="default"/> and <xref
elements. Other solutions can be classed as "partial TE" and also fall i target="RFC5394" format="default"/>. TE solutions may cover the
n mechanisms to distribute and/or enforce policies, but definition of
scope of this document.</t> specific policies is left to the network operator.</t>
<t>Path steering is the ability to forward packets using more
<t>Policy allows for the selection of paths (including next hops) based on information than just knowledge of the next hop. Examples of path
information beyond basic reachability. Early definitions of routing steering include IPv4 source routes <xref target="RFC0791"
policy, e.g., <xref target="RFC1102" /> and <xref target="RFC1104" />, format="default"/>, RSVP-TE explicit routes <xref target="RFC3209"
discuss routing policy being applied to restrict access to network format="default"/>, Segment Routing (SR) <xref target="RFC8402"
resources at an aggregate level. BGP is an example of a commonly used format="default"/>, and Service Function Chaining <xref
mechanism for applying such policies, see <xref target="RFC4271" /> and target="RFC7665" format="default"/>. Path steering for TE can be
<xref target="RFC8955" />. In the TE supported via control plane protocols, by encoding in the data plane
context, policy decisions are made within the control plane or by headers, or by a combination of the two. This includes when control
controllers in the management plane, and govern the selection of paths. is provided by a controller using a network-facing control
Examples can be found in <xref target="RFC4655" /> and <xref target="RFC5 protocol.</t>
394" />. <t>Resource management provides resource-aware control and forwarding.
TE solutions may cover the mechanisms to distribute and/or enforce Examples of resources are bandwidth, buffers, and queues, all of which
policies, but definition of specific policies is left to the network oper can be managed to control loss and latency.</t>
ator.</t> <t>Resource reservation is the control aspect of resource management.
It provides for domain-wide consensus about which network resources
<t>Path steering is the ability to forward packets using more information are used by a particular flow. This determination may be made at a
than just knowledge of the next hop. Examples of path steering include very coarse or very fine level. Note that this consensus exists at
IPv4 source routes <xref target="RFC0791" />, RSVP-TE explicit routes the network control or controller level but not within the data plane.
<xref target="RFC3209" />, Segment Routing <xref target="RFC8402" />, and It may be composed purely of accounting/bookkeeping, but it typically
Service Function Chaining <xref target="RFC7665" />. Path steering for T includes an ability to admit, reject, or reclassify a flow based on
E policy. Such accounting can be done based on any combination of a
can be supported via control plane protocols, by encoding in the data pla static understanding of resource requirements and the use of dynamic
ne mechanisms to collect requirements (e.g., via RSVP-TE <xref
headers, or by a combination of the two. This includes when control is target="RFC3209" format="default"/>) and resource availability (e.g.,
provided by a controller using a network-facing control protocol.</t> via OSPF extensions for GMPLS <xref target="RFC4203"
format="default"/>).</t>
<t>Resource management provides resource-aware control and forwarding. <t>Resource allocation is the data plane aspect of resource
Examples of resources are bandwidth, buffers, and queues, all of which ca management. It provides for the allocation of specific node and
n link resources to specific flows. Example resources include
be managed to control loss and latency. buffers, policing, and rate-shaping mechanisms that are typically
<list style="none"> supported via queuing. Resource allocation also includes the matching
<t>Resource reservation is the control aspect of resource management. of a flow (i.e.,
It provides for domain-wide consensus about which network resources flow classification) to a particular set of allocated resources.
are used by a particular flow. This determination may be made at a The method of flow classification and granularity of resource
very coarse or very fine level. Note that this consensus exists management is technology-specific. Examples include Diffserv with
at the network control or controller level, not within the data plan dropping and remarking <xref target="RFC4594" format="default"/>,
e. MPLS-TE <xref target="RFC3209" format="default"/>, GMPLS-based Label
It may be composed purely of accounting/bookkeeping, but it typicall Switched Paths (LSPs) <xref target="RFC3945" format="default"/>, as
y well as controller-based solutions <xref target="RFC8453"
includes an ability to admit, reject, or reclassify a flow based on format="default"/>. This level of resource control, while optional,
policy. Such accounting can be done based on any combination of a is important in networks that wish to support network congestion
static understanding of resource requirements, and the use of dynami management policies to control or regulate the offered traffic to
c deliver different levels of service and alleviate network
mechanisms to collect requirements (e.g., via <xref target="RFC3209" congestion problems. It is also important in networks that wish to con
/>) trol the
and resource availability (e.g., via <xref target="RFC4203" />).</t> latency experienced by specific traffic flows.</t>
<t>Resource allocation is the data plane aspect of resource management.
It
provides for the allocation of specific node and link resources to
specific flows. Example resources include buffers, policing, and
rate-shaping mechanisms that are typically supported via queuing. It
also includes the matching of a flow (i.e., flow classification) to a
particular set of allocated resources. The method of flow classifica
tion
and granularity of resource management is technology-specific. Examp
les
include Diffserv with dropping and remarking <xref target="RFC4594" /
>,
MPLS-TE <xref target="RFC3209" />, and GMPLS based label switched pat
hs
<xref target="RFC3945" />, as well as controller-based solutions
<xref target="RFC8453" />. This level of resource control, while opt
ional,
is important in networks that wish to support network congestion mana
gement policies
to control or regulate the offered traffic to deliver different level
s of
service and alleviate network congestion problems, or those networks
that wish to
control the latency experienced by specific traffic flows.</t>
</list></t>
</section>
<section anchor="SCOPE" title="Scope">
<t>The scope of this document is intra-domain TE because this is the practic
al
level of TE technology that exists in the Internet at the time of writing
.
That is, it describes TE within a given autonomous system in the
Internet. This document discusses concepts pertaining to intra-domain
traffic control, including such issues as routing control,
micro and macro resource allocation, and the control coordination
problems that arise consequently.</t>
<t>This document describes and characterizes techniques already in
use or in advanced development for Internet TE. The
way these techniques fit together is discussed and scenarios in which
they are useful will be identified.</t>
<t>Although the emphasis in this document is on intra-domain traffic
engineering, an overview of the high-level considerations pertaining
to inter-domain TE is provided in <xref target="INTER" />.
Inter-domain Internet TE is crucial to the performance
enhancement of the world-wide Internet infrastructure.</t>
<t>Whenever possible, relevant requirements from existing IETF documents
and other sources are incorporated by reference.</t>
</section>
<section anchor="TERMS" title="Terminology">
<t>This section provides terminology which is useful for Internet
TE. The definitions presented apply to this
document. These terms may have other meanings elsewhere.</t>
<t><list style="hanging"> </section>
<t hangText='Busy hour:'> <section anchor="SCOPE" numbered="true" toc="default">
A one hour period within a specified interval of time <name>Scope</name>
(typically 24 hours) in which the traffic load in a network <t>The scope of this document is intra-domain TE because this is the
or sub-network is greatest.</t> practical level of TE technology that exists in the Internet at the
<t hangText='Congestion:'> time of writing. That is, this document describes TE within a given
A state of a network resource in which the traffic incident AS in the Internet. This document discusses concepts
on the resource exceeds its output capacity over an interval pertaining to intra-domain traffic control, including such issues as
of time. A small amount of congestion may be beneficial to routing control, micro and macro resource allocation, and control
ensure that network resources are run at full capacity, and coordination problems that arise consequently.</t>
this may be particularly true at the network edge where it is <t>This document describes and characterizes techniques already in use
desirable to ensure that user traffic is served as much as or in advanced development for Internet TE. The way these techniques
possible. Within the network, if congestion is allowed to fit together is discussed and scenarios in which they are useful are
build (such as when input traffic exceeds output traffic in identified.</t>
a sustained way) it will have a negative effect on user <t>Although the emphasis in this document is on intra-domain traffic
traffic.</t> engineering, an overview of the high-level considerations pertaining
<t hangText='Congestion avoidance:'> to inter-domain TE is provided in <xref target="INTER"
An approach to congestion management that attempts to format="default"/>. Inter-domain Internet TE is crucial to the
obviate the occurrence of congestion. Chiefly relevant to performance enhancement of the world-wide Internet infrastructure.</t>
network congestion although may form a part of demand-side <t>Whenever possible, relevant requirements from existing IETF
congestion management.</t> documents and other sources are incorporated by reference.</t>
<t hangText='Congestion response:'> </section>
An approach to congestion management that attempts to remedy <section anchor="TERMS" numbered="true" toc="default">
congestion problems that have already occurred.</t> <name>Terminology</name>
<t hangText='Constraint-based routing:'> <t>This section provides terminology that is useful for Internet TE.
A class of routing protocols that take specified traffic The definitions presented apply to this document. These terms may
attributes, network constraints, and policy constraints into have other meanings elsewhere.</t>
account when making routing decisions. Constraint-based <dl newline="false" spacing="normal">
routing is applicable to traffic aggregates as well as <dt>Busy hour:</dt>
flows. It is a generalization of QoS-based routing.</t> <dd>A one-hour period within a specified interval of time (typically
<t hangText='Demand-side congestion management:'> 24 hours) in which the traffic load in a network or sub-network is
A congestion management scheme that addresses congestion greatest.</dd>
problems by regulating or conditioning offered load.</t> <dt>Congestion:</dt>
<t hangText='Effective bandwidth:'> <dd>A state of a network resource in which the traffic incident on
The minimum amount of bandwidth that can be assigned to a the resource exceeds its output capacity over an interval of time.
flow or traffic aggregate in order to deliver &apos;acceptable A small amount of congestion may be beneficial to ensure that
service quality&apos; to the flow or traffic aggregate. See network resources are run at full capacity, and this may be
<xref target="KELLY" /> for a more mathematical definition.</t> particularly true at the network edge where it is desirable to
<t hangText='Egress node:'> ensure that user traffic is served as much as possible. Within the
The device (router) at which traffic leaves a network toward a network, if congestion is allowed to build (such as when input
destination (host, server, etc.) or to another network.</t> traffic exceeds output traffic in a sustained way), it will have a
<t hangText='End-to-end:'> negative effect on user traffic.</dd>
This term is context-dependent and often applies to the life of <dt>Congestion avoidance:</dt>
a traffic flow from original source to final destination. In <dd>An approach to congestion management that attempts to obviate
contrast, edge-to-edge is often used to describe the traffic the occurrence of congestion. It is chiefly relevant to network
flow from the entry to a domain or network, to the exit from congestion, although it may form a part of demand-side congestion
that domain or network. In some contexts, however, for example management.</dd>
where there is a service interface between a network and the <dt>Congestion response:</dt>
client of that network, or where a path traverses multiple domains <dd>An approach to congestion management that attempts to remedy
under the control of a single process, end-to-end is used to refer congestion problems that have already occurred.</dd>
to the full operation of the service that may be composed of concaten <dt>Constraint-based routing:</dt>
ated <dd>A class of routing protocols that takes specified traffic
edge-to-edge operations. Thus, in the context of TE, the term end-to attributes, network constraints, and policy constraints into account
-end when making routing decisions. Constraint-based routing is
may refer to the full TE path, but not to the complete path of the tr applicable to traffic aggregates as well as flows. It is a
affic generalization of QoS-based routing.</dd>
from source application to ultimate destination.</t> <dt>Demand-side congestion management:</dt>
<t hangText='Hot-spot:'> <dd>A congestion management scheme that addresses congestion
A network element or subsystem which is in a considerably higher stat problems by regulating or conditioning the offered load.</dd>
e of <dt>Effective bandwidth:</dt>
congestion than others.</t> <dd>The minimum amount of bandwidth that can be assigned to a flow
<t hangText='Ingress node:'> or traffic aggregate in order to deliver "acceptable service
The device (router) at which traffic enters a network from a quality" to the flow or traffic aggregate. See <xref target="KELLY"
source (host) or from another network.</t> format="default"/> for a more mathematical definition.</dd>
<t hangText='Metric:'> <dt>Egress node:</dt>
A parameter defined in terms of standard units of <dd>The device (router) at which traffic leaves a network toward a
measurement.</t> destination (host, server, etc.) or to another network.</dd>
<t hangText='Measurement methodology:'> <dt>End-to-end:</dt>
<dd>This term is context-dependent and often applies to the life of
a traffic flow from original source to final destination.
In contrast, edge-to-edge is often used to describe the traffic flow from the
entry of a domain or network to the exit of that domain or network. However,
in some contexts (for example, where there is a service interface between a
network and the client of that network or where a path traverses multiple
domains under the control of a single process), end-to-end is used to refer to
the full operation of the service that may be composed of concatenated
edge-to-edge operations. Thus, in the context of TE, the term "end-to-end"
may refer to the full TE path but not to the complete path of the traffic from
source application to ultimate destination.</dd>
<dt>Hotspot:</dt>
<dd>A network element or subsystem that is in a considerably higher
state of congestion than others.</dd>
<dt>Ingress node:</dt>
<dd>The device (router) at which traffic enters a network from a
source (host) or from another network.</dd>
<dt>Metric:</dt>
<dd>A parameter defined in terms of standard units of
measurement.</dd>
<dt>Measurement methodology:</dt>
<dd>
A repeatable measurement technique used to derive one or A repeatable measurement technique used to derive one or
more metrics of interest.</t> more metrics of interest.</dd>
<t hangText='Network congestion:'> <dt>Network congestion:</dt>
Congestion within the network at a specific node or a <dd>
specific link that is sufficiently extreme that it results in Congestion within the network at a specific node or a specific link
unacceptable queuing delay or packet loss. Network congestion that is sufficiently extreme that it results in
can negatively impact end-to-end or edge-to-edge traffic flows, unacceptable queuing delay or packet loss. Network congestion can
so TE schemes may be deployed to balance traffic in the network negatively impact end-to-end or edge-to-edge traffic flows, so TE
and deliver congestion avoidance.</t> schemes may be deployed to balance traffic in the network and
<t hangText='Network survivability:'> deliver congestion avoidance.</dd>
<dt>Network survivability:</dt>
<dd>
The capability to provide a prescribed level of QoS for The capability to provide a prescribed level of QoS for
existing services after a given number of failures occur existing services after a given number of failures occur
within the network.</t> within the network.</dd>
<t hangText='Offered load:'> <dt>Offered load:</dt>
The offered load or offered traffic load is a measure of the <dd>Offered load is also sometimes called "offered traffic load". It
amount of traffic being presented to be carried across a is a measure of the amount of traffic being presented to be carried
network compared to the capacity of the network to carry it. across a network compared to the capacity of the network to carry
This term derives from queuing theory and an offered load of it. This term derives from queuing theory, and an offered load of 1
1 indicates that the network can carry, but only just manage indicates that the network can carry, but only just manage to carry,
to carry, all of the traffic presented to it.</t> all of the traffic presented to it.</dd>
<t hangText='Offline traffic engineering:'> <dt>Offline traffic engineering:</dt>
<dd>
A traffic engineering system that exists outside of the A traffic engineering system that exists outside of the
network.</t> network.</dd>
<t hangText='Online traffic engineering:'> <dt>Online traffic engineering:</dt>
A traffic engineering system that exists within the network, <dd>
A traffic-engineering system that exists within the network,
typically implemented on or as adjuncts to operational typically implemented on or as adjuncts to operational
network elements.</t> network elements.</dd>
<t hangText='Performance measures:'> <dt>Performance measures:</dt>
<dd>
Metrics that provide quantitative or qualitative measures of Metrics that provide quantitative or qualitative measures of
the performance of systems or subsystems of interest.</t> the performance of systems or subsystems of interest.</dd>
<t hangText='Performance metric:'> <dt>Performance metric:</dt>
<dd>
A performance parameter defined in terms of standard units A performance parameter defined in terms of standard units
of measurement.</t> of measurement.</dd>
<t hangText='Provisioning:'> <dt>Provisioning:</dt>
<dd>
The process of assigning or configuring network resources to The process of assigning or configuring network resources to
meet certain requests.</t> meet certain requests.</dd>
<t hangText='Quality of Service (QoS):'> <dt>Quality of Service (QoS):</dt>
QoS (<xref target="RFC3198" />) refers to the mechanisms used <dd>
within a network to achieve specific goals for the delivery of QoS <xref target="RFC3198" format="default"/> refers to the
traffic for a particular service according to the parameters mechanisms used within a network to achieve specific goals for the
specified in a Service Level Agreement. "Quality" is delivery of traffic for a particular service according to the
characterized by service availability, delay, jitter, throughput parameters specified in a Service Level Agreement. "Quality"
is characterized by service availability, delay, jitter, throughput,
and packet loss ratio. At a network resource level, "Quality of and packet loss ratio. At a network resource level, "Quality of
Service" refers to a set of capabilities that allow a service Service" refers to a set of capabilities that allow a service
provider to prioritize traffic, control bandwidth, and network provider to prioritize traffic, control bandwidth, and network
latency.</t> latency.</dd>
<t hangText='QoS routing:'> <dt>QoS routing:</dt>
<dd>
Class of routing systems that selects paths to be used by a Class of routing systems that selects paths to be used by a
flow based on the QoS requirements of the flow.</t> flow based on the QoS requirements of the flow.</dd>
<t hangText='Service Level Agreement (SLA):'> <dt>Service Level Agreement (SLA):</dt>
<dd>
A contract between a provider and a customer that guarantees A contract between a provider and a customer that guarantees
specific levels of performance and reliability at a certain specific levels of performance and reliability at a certain
cost.</t> cost.</dd>
<t hangText='Service Level Objective (SLO):'> <dt>Service Level Objective (SLO):</dt>
<dd>
A key element of an SLA between a provider and a customer. SLOs A key element of an SLA between a provider and a customer. SLOs
are agreed upon as a means of measuring the performance of the are agreed upon as a means of measuring the performance of the
Service Provider and are outlined as a way of avoiding disputes service provider and are outlined as a way of avoiding disputes
between the two parties based on misunderstanding.</t> between the two parties based on misunderstanding.</dd>
<t hangText='Stability:'> <dt>Stability:</dt>
<dd>
An operational state in which a network does not oscillate An operational state in which a network does not oscillate
in a disruptive manner from one mode to another mode.</t> in a disruptive manner from one mode to another mode.</dd>
<t hangText='Supply-side congestion management:'> <dt>Supply-side congestion management:</dt>
<dd>
A congestion management scheme that provisions additional A congestion management scheme that provisions additional
network resources to address existing and/or anticipated network resources to address existing and/or anticipated
congestion problems.</t> congestion problems.</dd>
<t hangText='Traffic characteristic:'> <dt>Traffic characteristic:</dt>
<dd>
A description of the temporal behavior or a description of A description of the temporal behavior or a description of
the attributes of a given traffic flow or traffic aggregate.</t> the attributes of a given traffic flow or traffic aggregate.</dd>
<t hangText='Traffic engineering system:'> <dt>Traffic-engineering system:</dt>
<dd>
A collection of objects, mechanisms, and protocols that are A collection of objects, mechanisms, and protocols that are
used together to accomplish traffic engineering objectives.</t> used together to accomplish traffic-engineering objectives.</dd>
<t hangText='Traffic flow:'> <dt>Traffic flow:</dt>
A stream of packets between two end-points that can be <dd>
A stream of packets between two endpoints that can be
characterized in a certain way. A common classification for a characterized in a certain way. A common classification for a
traffic flow selects packets with the "five-tuple" of source traffic flow selects packets with the five-tuple of source
and destination addresses, source and destination ports, and and destination addresses, source and destination ports, and
protocol ID. Flows may be very small and transient, ranging to protocol ID. Flows may be very small and transient, ranging to
very large. The TE techniques described in this document are very large. The TE techniques described in this document are
likely to be more effective when applied to large flows. Traffic likely to be more effective when applied to large flows. Traffic
flows may be aggregated and treated as a single unit in some flows may be aggregated and treated as a single unit in some
forms of TE making it possible to apply TE to the smaller flows forms of TE, making it possible to apply TE to the smaller flows
that comprise the aggregate.</t> that comprise the aggregate.</dd>
<t hangText='Traffic mapping:'> <dt>Traffic mapping:</dt>
Traffic mapping is the assignment of traffic workload onto (pre- <dd>
established) paths to meet certain requirements.</t> Traffic mapping is the assignment of traffic workload onto
<t hangText='Traffic matrix:'> (pre-established) paths to meet certain requirements.</dd>
<dt>Traffic matrix:</dt>
<dd>
A representation of the traffic demand between a set of A representation of the traffic demand between a set of
origin and destination abstract nodes. An abstract node can origin and destination abstract nodes. An abstract node can
consist of one or more network elements.</t> consist of one or more network elements.</dd>
<t hangText='Traffic monitoring:'> <dt>Traffic monitoring:</dt>
<dd>
The process of observing traffic characteristics at a given The process of observing traffic characteristics at a given
point in a network and collecting the traffic information point in a network and collecting the traffic information
for analysis and further action.</t> for analysis and further action.</dd>
<t hangText='Traffic trunk:'> <dt>Traffic trunk:</dt>
<dd>
An aggregation of traffic flows belonging to the same class An aggregation of traffic flows belonging to the same class
which are forwarded through a common path. A traffic trunk that are forwarded through a common path. A traffic trunk
may be characterized by an ingress and egress node, and a may be characterized by an ingress and egress node and a
set of attributes which determine its behavioral set of attributes that determine its behavioral
characteristics and requirements from the network.</t> characteristics and requirements from the network.</dd>
<t hangText='Workload:'> <dt>Workload:</dt>
The workload or traffic workload is an evaluation of the amount <dd>Workload is also sometimes called "traffic workload". It is an
of work that must be done in a network in order to facilitate evaluation of the amount of work that must be done in a network in
the traffic demand. Colloquially, it is the answer to, "How order to facilitate the traffic demand. Colloquially, it is the
busy is the network?"</t> answer to, "How busy is the network?"</dd>
</list></t> </dl>
</section>
</section> </section>
<section anchor="BG" numbered="true" toc="default">
</section> <name>Background</name>
<t>The Internet aims to convey IP packets from ingress nodes to egress nod
<section anchor="BG" title="Background"> es
<t>The Internet aims to convey IP packets from ingress nodes to egress nodes
efficiently, expeditiously, and economically. Furthermore, in a efficiently, expeditiously, and economically. Furthermore, in a
multiclass service environment (e.g., Diffserv capable networks - see multi-class service environment (e.g., Diffserv capable networks; see
<xref target="DIFFSERV" />), the resource sharing parameters of the <xref target="DIFFSERV" format="default"/>), the resource-sharing parameter
s of the
network must be appropriately determined and configured according to network must be appropriately determined and configured according to
prevailing policies and service models to resolve resource contention prevailing policies and service models to resolve resource contention
issues arising from mutual interference between packets traversing issues arising from mutual interference between packets traversing
the network. Thus, consideration must be given to resolving the network. Thus, consideration must be given to resolving
competition for network resources between traffic flows belonging to competition for network resources between traffic flows belonging to
the same service class (intra-class contention resolution) and traffic the same service class (intra-class contention resolution) and traffic
flows belonging to different classes (inter-class contention flows belonging to different classes (inter-class contention
resolution).</t> resolution).</t>
<section anchor="CONTEXT" numbered="true" toc="default">
<section anchor="CONTEXT" title="Context of Internet Traffic Engineering"> <name>Context of Internet Traffic Engineering</name>
<t>The context of Internet traffic engineering includes the following su
<t>The context of Internet traffic engineering includes the following sub-co b-contexts:</t>
ntexts:</t> <ol spacing="normal" type="1">
<li>A network domain context that defines the scope under
<t><list style="numbers"> consideration and, in particular, the situations in which the TE
<t>A network domain context that defines the scope under consideration, problems occur. The network domain context includes network
and in particular the situations in which the TE structure, policies, characteristics, constraints, quality
problems occur. The network domain context includes network attributes, and optimization criteria.</li>
structure, policies, characteristics, constraints, quality attribute <li>A problem context defining the general and concrete issues that
s, TE addresses. The problem context includes identification,
and optimization criteria.</t> abstraction of relevant features, representation, formulation,
<t>A problem context defining the general and concrete issues specification of the requirements on the solution space, and
that TE addresses. The problem context specification of the desirable features of acceptable
includes identification, abstraction of relevant features, solutions.</li>
representation, formulation, specification of the <li>A solution context suggesting how to address the issues
requirements on the solution space, and specification of the identified by the problem context. The solution context includes
desirable features of acceptable solutions.</t> analysis, evaluation of alternatives, prescription, and
<t>A solution context suggesting how to address the issues resolution.</li>
identified by the problem context. The solution context <li>An implementation and operational context in which the
includes analysis, evaluation of alternatives, prescription,
and resolution.</t>
<t>An implementation and operational context in which the
solutions are instantiated. The implementation and operational solutions are instantiated. The implementation and operational
context includes planning, organization, and execution.</t> context includes planning, organization, and execution.</li>
</list></t> </ol>
<t>The context of Internet TE and the different problem
<t>The context of Internet TE and the different problem
scenarios are discussed in the following subsections.</t> scenarios are discussed in the following subsections.</t>
</section>
</section> <section anchor="NWCTXT" numbered="true" toc="default">
<name>Network Domain Context</name>
<section anchor="NWCTXT" title="Network Domain Context"> <t>IP networks range in size from small clusters of routers situated
within a given location to thousands of interconnected routers,
<t>IP networks range in size from small clusters of routers situated switches, and other components distributed all over the world.</t>
within a given location, to thousands of interconnected routers, <t>At the most basic level of abstraction, an IP network can be
switches, and other components distributed all over the world.</t> represented as a distributed dynamic system consisting of:
</t>
<t>At the most basic level of abstraction, an IP network can be represented <ul spacing="normal">
as a distributed dynamic system consisting of: <li>a set of interconnected resources that provide transport
<list style="symbols"> services for IP traffic subject to certain constraints</li>
<t>a set of interconnected resources which provide transport <li>a demand system representing the offered load to be transported
services for IP traffic subject to certain constraints</t> through the network</li>
<t>a demand system representing the offered load to be transported <li>a response system consisting of network processes, protocols,
through the network</t> and related mechanisms that facilitate the movement of traffic
<t>a response system consisting of network processes, protocols, through the network (see also <xref target="AWD2"
and related mechanisms which facilitate the movement of format="default"/>)</li>
traffic through the network (see also <xref target="AWD2"/>).</t> </ul>
</list></t> <t>The network elements and resources may have specific
characteristics restricting the manner in which the traffic demand is
<t>The network elements and resources may have specific characteristics handled. Additionally, network resources may be equipped with traffic
restricting the manner in which the traffic demand is handled. Additiona control mechanisms managing the way in which the demand is serviced.
lly, Traffic control mechanisms may be used to:
network resources may be equipped with traffic control mechanisms managin </t>
g <ul spacing="normal">
the way in which the demand is serviced. Traffic control mechanisms may <li>control packet processing activities within a given
be resource</li>
used to: <li>arbitrate contention for access to the resource by different
<list style="symbols"> packets</li>
<t>control packet processing activities within a given resource</t> <li>regulate traffic behavior through the resource</li>
<t>arbitrate contention for access to the resource by different </ul>
packets</t> <t>A configuration management and provisioning system may allow the
<t>regulate traffic behavior through the resource.</t> settings of the traffic control mechanisms to be manipulated by
</list></t> external or internal entities in order to exercise control over the
way in which the network elements respond to internal and external
<t>A configuration management and provisioning system may allow the stimuli.</t>
settings of the traffic control mechanisms to be manipulated by <t>The details of how the network carries packets are specified in the
external or internal entities in order to exercise control over the policies of the network administrators and are installed through
way in which the network elements respond to internal and external network configuration management and policy-based provisioning
stimuli.</t> systems. Generally, the types of service provided by the network also
depend upon the technology and characteristics of the network elements
<t>The details of how the network carries packets are specified in the and protocols, the prevailing service and utility models, and the
policies of the network administrators and are installed through ability of the network administrators to translate policies into
network configuration management and policy-based provisioning systems. network configurations.</t>
Generally, the types of service provided by the network also depend <t>Internet networks have two significant characteristics:</t>
upon the technology and characteristics of the network elements and <ul spacing="normal">
protocols, the prevailing service and utility models, and the ability <li>They provide real-time services.</li>
of the network administrators to translate policies into network <li>Their operating environments are very dynamic.</li>
configurations.</t> </ul>
<t>The dynamic characteristics of IP and IP/MPLS networks can be
<t>Internet networks have two significant characteristics: attributed in part to fluctuations in demand, the interaction between
<list style="symbols"> various network protocols and processes, the rapid evolution of the
<t>they provide real-time services</t> infrastructure that demands the constant inclusion of new technologies
<t>their operating environments are very dynamic.</t> and new network elements, and the transient and persistent faults that
</list></t> occur within the system.</t>
<t>Packets contend for the use of network resources as they are
<t>The dynamic characteristics of IP and IP/MPLS networks can be attributed conveyed through the network. A network resource is considered to be
in part congested if, for an interval of time, the arrival rate of packets
to fluctuations in demand, to the interaction between various network exceeds the output capacity of the resource. Network congestion may
protocols and processes, to the rapid evolution of the infrastructure result in some of the arriving packets being delayed or even
which demands the constant inclusion of new technologies and new network dropped.</t>
elements, and to transient and persistent faults which occur within <t>Network congestion increases transit delay and delay variation, may
the system.</t> lead to packet loss, and reduces the predictability of network
services. Clearly, while congestion may be a useful tool at ingress
<t>Packets contend for the use of network resources as they are conveyed edge nodes, network congestion is highly undesirable. Combating
through the network. A network resource is considered to be network congestion at a reasonable cost is a major objective of
congested if, for an interval of time, the arrival rate of packets Internet TE, although it may need to be traded with other objectives to
exceeds the output capacity of the resource. Network congestion may resu keep the costs reasonable.</t>
lt in <t>Efficient sharing of network resources by multiple traffic flows is
some of the arriving packets being delayed or even dropped.</t> a basic operational premise for the Internet. A fundamental challenge
in network operation is to increase resource utilization while
<t>Network congestion increases transit delay and delay variation, may lead minimizing the possibility of congestion.</t>
to packet loss, <t>The Internet has to function in the presence of different classes
and reduces the predictability of network services. Clearly, while conge of traffic with different service requirements. This requirement is
stion may clarified in the architecture for Differentiated Services (Diffserv)
be a useful tool at ingress edge nodes, network congestion is highly unde <xref target="RFC2475" format="default"/>. That document describes
sirable. how packets can be grouped into behavior aggregates such that each
Combating network congestion at a reasonable cost is a major objective of aggregate has a common set of behavioral characteristics or a common
Internet TE set of delivery requirements. Delivery requirements of a specific set
although it may need to be traded with other objectives to keep the costs of packets may be specified explicitly or implicitly. Two of the most
reasonable.</t> important traffic delivery requirements are:</t>
<ul spacing="normal">
<t>Efficient sharing of network resources by multiple traffic flows is <li>Capacity constraints can be expressed statistically as peak
a basic operational premise for the Internet. A fundamental challenge rates, mean rates, burst sizes, or as some deterministic notion of
in network operation is to increase resource utilization while minimizing effective bandwidth.</li>
the possibility of congestion.</t> <li><t>QoS requirements can be expressed in terms of:</t>
<ul spacing="normal">
<t>The Internet has to function in the presence of different classes of <li>integrity constraints, such as packet loss</li>
traffic with different service requirements. This requirement is <li>temporal constraints, such as timing restrictions for the
clarified in the architecture for Differentiated Services (Diffserv) <xre delivery of each packet (delay) and timing restrictions for the
f target="RFC2475" />. delivery of consecutive packets belonging to the same traffic
That document describes how packets can be grouped into behavior aggregat stream (delay variation)</li>
es such that </ul>
each aggregate has a common set of behavioral characteristics or a common </li>
set of delivery </ul>
requirements. Delivery requirements of a specific set of packets may </section>
be specified explicitly or implicitly. Two of the most important <section anchor="PRBCTXT" numbered="true" toc="default">
traffic delivery requirements are: <name>Problem Context</name>
<t>There are several problems associated with operating a
<list style="symbols"> network like those described in the previous section. This section analy
zes
<t>Capacity constraints can be expressed statistically as peak rates,
mean rates, burst sizes, or as some deterministic notion of effecti
ve
bandwidth.</t>
<t>QoS requirements can be expressed in terms of:
<list style="symbols">
<t>integrity constraints such as packet loss</t>
<t>temporal constraints such as timing restrictions for the deliv
ery
of each packet (delay) and timing restrictions for the deliver
y of
consecutive packets belonging to the same traffic stream (dela
y
variation).</t>
</list></t>
</list></t>
</section>
<section anchor="PRBCTXT" title="Problem Context">
<t>There are several problems associated with operating a
network described in the previous section. This section analyzes
the problem context in relation to TE. The the problem context in relation to TE. The
identification, abstraction, representation, and measurement of identification, abstraction, representation, and measurement of
network features relevant to TE are significant network features relevant to TE are significant
issues.</t> issues.</t>
<t>A particular challenge is to formulate the problems that traffic
<t>A particular challenge is to formulate the problems that traffic
engineering attempts to solve. For example: engineering attempts to solve. For example:
<list style="symbols"> </t>
<t>How to identify the requirements on the solution space?</t>
<t>How to specify the desirable features of solutions?</t>
<t>How to actually solve the problems?</t>
<t>How to measure and characterize the effectiveness of
solutions?</t>
</list></t>
<t>Another class of problems is how to measure and estimate <ul spacing="normal">
<li>How to identify the requirements on the solution space</li>
<li>How to specify the desirable features of solutions</li>
<li>How to actually solve the problems</li>
<li>How to measure and characterize the effectiveness of
solutions</li>
</ul>
<t>Another class of problems is how to measure and estimate
relevant network state parameters. Effective TE relevant network state parameters. Effective TE
relies on a good estimate of the offered traffic load as well as a relies on a good estimate of the offered traffic load as well as a
view of the underlying topology and associated resource constraints. view of the underlying topology and associated resource constraints.
Offline planning requires a full view of the topology of the network Offline planning requires a full view of the topology of the network
or partial network that is being planned.</t> or partial network that is being planned.</t>
<t>Still another class of problem is how to characterize the state of
<t>Still another class of problem is how to characterize the the network and how to evaluate its performance. The performance
state of the network and how to evaluate its performance. The evaluation problem is two-fold: one aspect relates to the evaluation
performance evaluation problem is two-fold: one aspect relates to of the system-level performance of the network, and the other aspect
the evaluation of the system-level performance of the network; the relates to the evaluation of resource-level performance, which
other aspect relates to the evaluation of resource-level performance, restricts attention to the performance analysis of individual network
which restricts attention to the performance analysis of individual resources.</t>
network resources.</t> <t>In this document, we refer to the system-level characteristics of
the network as the "macro-states" and the resource-level
<t>In this document, we refer to the system-level characteristics of the characteristics as the "micro-states." The system-level
network as the "macro-states" and the resource-level characteristics characteristics are also known as the emergent properties of the
as the "micro-states." The system-level characteristics are also known network. Correspondingly, we refer to the TE schemes dealing with
as the emergent properties of the network. Correspondingly, we refer network performance optimization at the systems level as "macro-TE"
to the TE schemes dealing with network performance and the schemes that optimize at the individual resource level as
optimization at the systems level as "macro-TE" and the schemes that "micro-TE." Under certain circumstances, the system-level performance
optimize at the individual resource level as "micro-TE." Under can be derived from the resource-level performance using appropriate
certain circumstances, the system-level performance can be derived rules of composition, depending upon the particular performance
from the resource-level performance using appropriate rules of measures of interest.</t>
composition, depending upon the particular performance measures of <t>Another fundamental class of problem concerns how to effectively
interest.</t>
<t>Another fundamental class of problem concerns how to effectively
optimize network performance. Performance optimization may entail optimize network performance. Performance optimization may entail
translating solutions for specific TE problems into translating solutions for specific TE problems into
network configurations. Optimization may also entail some degree of network configurations. Optimization may also entail some degree of
resource management control, routing control, and capacity resource management control, routing control, and capacity
augmentation.</t> augmentation.</t>
<section anchor="CONGEST" numbered="true" toc="default">
<section anchor="CONGEST" title="Congestion and its Ramifications"> <name>Congestion and Its Ramifications</name>
<t>Network congestion is one of the most significant problems in an op
<t>Network congestion is one of the most significant problems in an operat erational
ional
IP context. A network element is said to be congested if it IP context. A network element is said to be congested if it
experiences sustained overload over an interval of time. Although cong estion experiences sustained overload over an interval of time. Although cong estion
at the edge of the network may be beneficial in ensuring that the netwo rk at the edge of the network may be beneficial in ensuring that the netwo rk
delivers as much traffic as possible, network congestion delivers as much traffic as possible, network congestion
almost always results in degradation of service quality to end users. almost always results in degradation of service quality to end users.
Congestion avoidance and response schemes can include demand-side polic ies and Congestion avoidance and response schemes can include demand-side polic ies and
supply-side policies. Demand-side policies may restrict access to supply-side policies. Demand-side policies may restrict access to
congested resources or dynamically regulate the demand to alleviate congested resources or dynamically regulate the demand to alleviate
the overload situation. Supply-side policies may expand or augment the overload situation. Supply-side policies may expand or augment
network capacity to better accommodate offered traffic. Supply-side network capacity to better accommodate offered traffic. Supply-side
policies may also re-allocate network resources by redistributing policies may also reallocate network resources by redistributing
traffic over the infrastructure. Traffic redistribution and resource traffic over the infrastructure. Traffic redistribution and resource
re-allocation serve to increase the 'effective capacity' of the network reallocation serve to increase the effective capacity of the network.</
.</t> t>
<t>The emphasis of this document is primarily on congestion management
<t>The emphasis of this document is primarily on congestion management
schemes falling within the scope of the network, rather than on schemes falling within the scope of the network, rather than on
congestion management systems dependent upon sensitivity and congestion management systems dependent upon sensitivity and
adaptivity from end-systems. That is, the aspects that are adaptivity from end systems. That is, the aspects that are
considered in this document with respect to congestion management are considered in this document with respect to congestion management are
those solutions that can be provided by control entities operating on those solutions that can be provided by control entities operating on
the network and by the actions of network administrators and network the network and by the actions of network administrators and network
operations systems.</t> operations systems.</t>
</section>
</section>
<section anchor="SLNCTXT" numbered="true" toc="default">
<name>Solution Context</name>
<t>The solution context for Internet TE involves analysis,
evaluation of alternatives, and choice between alternative courses
of action. Generally, the solution context is based on making
inferences about the current or future state of the network and making
decisions that may involve a preference between alternative sets of
action. More specifically, the solution context demands reasonable
estimates of traffic workload, characterization of network state,
derivation of solutions that may be implicitly or explicitly
formulated, and possibly instantiation of a set of control
actions. Control actions may involve the manipulation of parameters
associated with routing, control over tactical capacity acquisition,
and control over the traffic management functions.</t>
<t>The following list of instruments may be applicable to the solution
context of Internet TE:</t>
<ul spacing="normal">
<li>A set of policies, objectives, and requirements (which may be
context dependent) for network performance evaluation and
performance optimization.</li>
<li>A collection of online and, in some cases, possibly offline tools
and mechanisms for measurement, characterization, modeling,
control of traffic, control over the placement and allocation of
network resources, as well as control over the mapping or
distribution of traffic onto the infrastructure.</li>
<li>A set of constraints on the operating environment, the network
protocols, and the TE system itself.</li>
<li>A set of quantitative and qualitative techniques and
methodologies for abstracting, formulating, and solving TE
problems.</li>
<li>A set of administrative control parameters that may be
manipulated through a configuration management system. Such a
system may itself include a configuration control subsystem, a
configuration repository, a configuration accounting subsystem, and
a configuration auditing subsystem.</li>
<li>A set of guidelines for network performance evaluation,
performance optimization, and performance improvement.</li>
</ul>
<t>Determining traffic characteristics through measurement or
estimation is very useful within the realm of the TE solution space.
Traffic estimates can be derived from customer subscription
information, traffic projections, traffic models, and actual
measurements. The measurements may be performed at different levels,
e.g., at the traffic-aggregate level or at the flow level.
Measurements at the flow level or on small traffic aggregates may be
performed at edge nodes, when traffic enters and leaves the network.
Measurements for large traffic aggregates may be performed within the
core of the network.</t>
<t>To conduct performance studies and to support planning of existing
and future networks, a routing analysis may be performed to determine
the paths the routing protocols will choose for various traffic
demands and to ascertain the utilization of network resources as
traffic is routed through the network. Routing analysis captures the
selection of paths through the network, the assignment of traffic
across multiple feasible routes, and the multiplexing of IP traffic
over traffic trunks (if such constructs exist) and over the underlying
network infrastructure. A model of network topology is necessary to
perform routing analysis. A network topology model may be extracted
from:
</t>
<ul spacing="normal">
<li>network architecture documents</li>
<li>network designs</li>
<li>information contained in router configuration files</li>
<li>routing databases such as the link-state database of an Interior
Gateway Protocol (IGP)</li>
<li>routing tables</li>
<li>automated tools that discover and collate network topology
information</li>
</ul>
<t>Topology information may also be derived from servers that monitor
network state and from servers that perform provisioning
functions.</t>
<t>Routing in operational IP networks can be administratively
controlled at various levels of abstraction, including the manipulation
of BGP attributes and IGP metrics. For path-oriented technologies
such as MPLS, routing can be further controlled by the manipulation of
relevant TE parameters, resource parameters, and administrative policy
constraints. Within the context of MPLS, the path of an explicitly
routed LSP can be computed and established in
various ways, including:
</t>
<ul spacing="normal">
<li>manually</li>
<li>automatically and online using constraint-based routing processes
implemented on Label Switching Routers (LSRs)</li>
<li>automatically and offline using constraint-based routing entities
implemented on external TE support systems</li>
</ul>
<section anchor="COMBAT" numbered="true" toc="default">
<name>Combating the Congestion Problem</name>
<t>Minimizing congestion is a significant aspect of Internet traffic
engineering. This subsection gives an overview of the general
approaches that have been used or proposed to combat congestion.</t>
<t>Congestion management policies can be categorized based upon the
following criteria (see <xref target="YARE95" format="default"/> for
a more detailed taxonomy of congestion control schemes):</t>
<ol spacing="normal" type="1"><li>
<t>Congestion Management Based on Response Timescales </t>
<ul spacing="normal">
<li>Long (weeks to months): Expanding network capacity by
adding new equipment, routers, and links takes time and is
comparatively costly. Capacity planning needs to take this
into consideration. Network capacity is expanded based on
estimates or forecasts of future traffic development and
traffic distribution. These upgrades are typically carried
out over weeks, months, or maybe even years.</li>
<li><t>Medium (minutes to days): Several control policies fall
within the medium timescale category. Examples include:</t>
<ol spacing="normal" type="a">
<li>Adjusting routing protocol parameters to route traffic
away from or towards certain segments of the network.</li>
<li>Setting up or adjusting explicitly routed LSPs in MPLS
networks to route traffic trunks away from possibly
congested resources or toward possibly more favorable
routes.</li>
<li>Reconfiguring the logical topology of the network to
make it correlate more closely with the spatial traffic
distribution using, for example, an underlying
path-oriented technology such as MPLS LSPs or optical
channel trails.</li>
</ol>
<t>When these schemes are adaptive, they rely on measurement
systems. A measurement system monitors changes in traffic
distribution, traffic loads, and network resource
utilization and then provides feedback to the online or
offline TE mechanisms and tools so that they can trigger
control actions within the network. The TE mechanisms and
tools can be implemented in a distributed or centralized
fashion. A centralized scheme may have full visibility into
the network state and may produce more optimal solutions.
However, centralized schemes are prone to single points of
failure and may not scale as well as distributed schemes.
Moreover, the information utilized by a centralized scheme
may be stale and might not reflect the actual state of the
network. It is not an objective of this document to make a
recommendation between distributed and centralized schemes;
that is a choice that network administrators must make based
on their specific needs.</t>
</li>
<li>Short (minutes or less): This category includes
packet-level processing functions and events that are recorded
on the order of several round-trip times. It also includes
router mechanisms such as passive and active buffer
management. All of these mechanisms are used to control
congestion or signal congestion to end systems so that they
can adaptively regulate the rate at which traffic is injected
into the network. A well-known active queue management
scheme, especially for responsive traffic such as TCP, is
Random Early Detection (RED) <xref target="FLJA93"
format="default"/>. During congestion (but before the queue
is filled), the RED scheme chooses arriving packets to "mark"
according to a probabilistic algorithm that takes into account
the average queue size. A router that does not utilize
Explicit Congestion Notification (ECN) <xref target="RFC3168"
format="default"/> can simply drop marked packets to alleviate
congestion and implicitly notify the receiver about the
congestion. On the other hand, if the router and the end
hosts support ECN, they can set the ECN field in the packet
header, and the end host can act on this information. Several
variations of RED have been proposed to support different drop
precedence levels in multi-class environments <xref
target="RFC2597" format="default"/>. RED provides congestion
avoidance that is better than or equivalent to Tail-Drop (TD)
queue management (drop arriving packets only when the queue is
full). Importantly, RED reduces the possibility of
retransmission bursts becoming synchronized within the network
and improves fairness among different responsive traffic
sessions. However, RED by itself cannot prevent congestion
and unfairness caused by sources unresponsive to RED, e.g.,
some misbehaved greedy connections. Other schemes have been
proposed to improve performance and fairness in the presence
of unresponsive traffic. Some of those schemes (such as
Longest Queue Drop (LQD) and Dynamic Soft Partitioning with
Random Drop (RND) <xref target="SLDC98" format="default"/>)
were proposed as theoretical frameworks and are typically not
available in existing commercial products, while others (such
as Approximate Fair Dropping (AFD) <xref target="AFD03"
format="default"/>) have seen some implementation. Advice on
the use of Active Queue Management (AQM) schemes is provided
in <xref target="RFC7567" format="default"/>. <xref
target="RFC7567" format="default"/> recommends self-tuning AQM
algorithms like those that the IETF has published in <xref
target="RFC8290" format="default"/>, <xref target="RFC8033"
format="default"/>, <xref target="RFC8034" format="default"/>,
and <xref target="RFC9332" format="default"/>, but RED is
still appropriate for links with stable bandwidth, if
configured carefully.</li>
</ul>
</li>
<li><t>Reactive versus Preventive Congestion Management Schemes
</t>
<ul spacing="normal">
<li>Reactive (recovery) congestion management policies react
to existing congestion problems. All the policies described
above for the short and medium timescales can be categorized
as being reactive. They are based on monitoring and
identifying congestion problems that exist in the network and
on the initiation of relevant actions to ease a situation.
Reactive congestion management schemes may also be
preventive.</li>
<li>Preventive (predictive/avoidance) policies take proactive
action to prevent congestion based on estimates and
predictions of future congestion problems (e.g., traffic
matrix forecasts). Some of the policies described for the
long and medium timescales fall into this category.
Preventive policies do not necessarily respond immediately to
existing congestion problems. Instead, forecasts of traffic
demand and workload distribution are considered, and action
may be taken to prevent potential future congestion problems.
The schemes described for the short timescale can also be used
for congestion avoidance because dropping or marking packets
before queues actually overflow would trigger corresponding
responsive traffic sources to slow down. Preventive
congestion management schemes may also be reactive.</li>
</ul>
</li>
<li><t>Supply-Side versus Demand-Side Congestion Management
Schemes</t>
<ul spacing="normal">
<li>Supply-side congestion management policies increase the
effective capacity available to traffic in order to control or
reduce congestion. This can be accomplished by increasing
capacity or by balancing distribution of traffic over the
network.
Capacity planning aims to provide a physical
topology and associated link bandwidths that match or exceed
estimated traffic workload and traffic distribution, subject to
traffic forecasts and budgetary (or other) constraints. If the
actual traffic distribution does not fit the topology derived
from capacity planning, then the traffic can be mapped onto
the topology by using routing control mechanisms, by applying
path-oriented technologies (e.g., MPLS LSPs and optical
channel trails) to modify the logical topology or by
employing some other load redistribution mechanisms.</li>
<li>Demand-side congestion management policies control or
regulate the offered traffic to alleviate congestion problems.
For example, some of the short timescale mechanisms described
earlier as well as policing and rate-shaping mechanisms
attempt to regulate the offered load in various ways.</li>
</ul>
</li>
</ol>
</section> </section>
</section>
</section> <section anchor="IMPCTXT" numbered="true" toc="default">
<name>Implementation and Operational Context</name>
<section anchor="SLNCTXT" title="Solution Context"> <t>The operational context of Internet TE is
<t>The solution context for Internet TE involves
analysis, evaluation of alternatives, and choice between alternative
courses of action. Generally, the solution context is based on
making inferences about the current or future state of the
network, and making decisions that may involve a preference between
alternative sets of action. More specifically, the solution context
demands reasonable estimates of traffic workload, characterization of
network state, derivation of solutions which may be implicitly or
explicitly formulated, and possibly instantiating a set of control
actions. Control actions may involve the manipulation of parameters
associated with routing, control over tactical capacity acquisition,
and control over the traffic management functions.</t>
<t>The following list of instruments may be applicable to the solution
context of Internet TE.</t>
<t><list style="symbols">
<t>A set of policies, objectives, and requirements (which may
be context dependent) for network performance evaluation and
performance optimization.</t>
<t>A collection of online and in some cases possibly offline tools and m
echanisms
for measurement, characterization, modeling, and control of traffic,
and control over the placement and allocation of network resources,
as well as control over the mapping or distribution of traffic onto
the infrastructure.</t>
<t>A set of constraints on the operating environment, the network
protocols, and the TE system itself.</t>
<t>A set of quantitative and qualitative techniques and methodologies
for abstracting, formulating, and solving TE
problems.</t>
<t>A set of administrative control parameters which may be
manipulated through a configuration management system. Such a
system may, itself, include a configuration control subsystem,
a configuration repository, a configuration accounting subsystem,
and a configuration auditing subsystem.</t>
<t>A set of guidelines for network performance evaluation,
performance optimization, and performance improvement.</t>
</list></t>
<t>Determining traffic characteristics through measurement or
estimation is very useful within the realm of the TE
solution space. Traffic estimates can be derived from customer
subscription information, traffic projections, traffic models, and
from actual measurements. The measurements may be performed at
different levels, e.g., at the traffic-aggregate level or at the flow
level. Measurements at the flow level or on small traffic aggregates
may be performed at edge nodes, when traffic enters and leaves the
network. Measurements for large traffic-aggregates may be performed
within the core of the network.</t>
<t>To conduct performance studies and to support planning of existing
and future networks, a routing analysis may be performed to determine
the paths the routing protocols will choose for various traffic
demands, and to ascertain the utilization of network resources as
traffic is routed through the network. Routing analysis captures the
selection of paths through the network, the assignment of traffic across
multiple feasible routes, and the multiplexing of IP traffic over traffic
trunks (if such constructs exist) and over the underlying network
infrastructure. A model of network topology is necessary to perform
routing analysis. A network topology model may be extracted from:
<list style="symbols">
<t>network architecture documents</t>
<t>network designs</t>
<t>information contained in router configuration files</t>
<t>routing databases such as the link state database of an interior
gateway protocol (IGP)</t>
<t>routing tables</t>
<t>automated tools that discover and collate network topology
information.</t>
</list></t>
<t>Topology information may also be derived from servers that monitor
network state, and from servers that perform provisioning functions.</t>
<t>Routing in operational IP networks can be administratively controlled
at various levels of abstraction including the manipulation of BGP
attributes and IGP metrics. For path-oriented
technologies such as MPLS, routing can be further controlled by the manip
ulation
of relevant TE parameters, resource parameters, and administrative
policy constraints. Within the context of MPLS, the path of an explicitl
y
routed label switched path (LSP) can be computed and established in vario
us
ways including:
<list style="symbols">
<t>manually</t>
<t>automatically, online using constraint-based routing processes
implemented on label switching routers</t>
<t>automatically, offline using constraint-based routing entities
implemented on external TE support systems.</t>
</list></t>
<section anchor="COMBAT" title="Combating the Congestion Problem">
<t>Minimizing congestion is a significant aspect of Internet traffic
engineering. This subsection gives an overview of the general
approaches that have been used or proposed to combat congestion.</t>
<t>Congestion management policies can be categorized based upon the
following criteria (see <xref target="YARE95" /> for a more
detailed taxonomy of congestion control schemes):
<list style="numbers">
<t>Congestion Management Based on Response Timescales
<list style="symbols">
<t>Long (weeks to months): Expanding network capacity by adding
new
equipment, routers, and links takes time and is comparatively
costly. Capacity planning needs to take this into considerat
ion.
Network capacity is expanded based on estimates or forecasts
of
future traffic development and traffic distribution. These
upgrades are typically carried out over weeks or months, or m
aybe
even years.</t>
<t>Medium (minutes to days): Several control policies fall withi
n the
medium timescale category. Examples include:
<list style="letters">
<t>Adjusting routing protocol parameters to route traffic a
way from or
towards certain segments of the network.</t>
<t>Setting up or adjusting explicitly routed LSPs in MPLS
networks to route traffic trunks away from possibly
congested resources or toward possibly more favorable ro
utes.</t>
<t>Re-configuring the logical topology of the network to ma
ke it
correlate more closely with the spatial traffic distribu
tion
using, for example, an underlying path-oriented technolo
gy such
as MPLS LSPs or optical channel trails.</t>
</list>
When these schemes are adaptive, they rely on measurement sys
tems. A measurement
system monitors changes in traffic distribution, traffic load
s, and network
resource utilization and then provides feedback to the online
or offline
TE mechanisms and tools so that they can trigger control
actions within the network. The TE mechanisms and tools
can be implemented in a distributed or centralized fashion.
A centralized
scheme may have full visibility into the network state and ma
y produce
more optimal solutions. However, centralized schemes are pro
ne to single
points of failure and may not scale as well as distributed sc
hemes. Moreover,
the information utilized by a centralized scheme may be stale
and might not
reflect the actual state of the network. It is not an object
ive of this
document to make a recommendation between distributed and cen
tralized
schemes: that is a choice that network administrators must ma
ke based on
their specific needs.</t>
<t>Short (minutes or less): This category includes packet level
processing functions and events that are recorded on the orde
r of several
round-trip times. It also includes router mechanisms such as
passive and
active buffer management. All of these mechanisms are used t
o control
congestion or signal congestion to end systems so that they c
an adaptively
regulate the rate at which traffic is injected into the netwo
rk. A well-known
active queue management scheme, especially for
responsive traffic such as TCP, is Random Early Detection (RE
D) <xref target="FLJA93"/>.
During congestion (but before the queue is filled), the RED s
cheme chooses
arriving packets to "mark" according to a probabilistic algor
ithm
which takes into account the average queue size. A router th
at does not
utilize explicit congestion notification (ECN) <xref target="
RFC3168" /> can
simply drop marked packets to alleviate congestion and implic
itly notify the
receiver about the congestion. On the other hand, if the rou
ter and the end hosts support ECN,
they can set the ECN field in the packet header, and the end
host can act on this
information. Several variations of RED have
been proposed to support different drop precedence levels in
multi-class
environments <xref target="RFC2597"/>. RED provides congesti
on avoidance
which is better than or equivalent to traditional Tail-Drop (
TD) queue management (drop
arriving packets only when the queue is full). Importantly,
RED reduces the
possibility of retransmission bursts becoming synchronized wi
thin the network,
and improves fairness among different
responsive traffic sessions. However, RED by itself cannot p
revent congestion and unfairness
caused by sources unresponsive to RED, e.g., some misbehaved
greedy connections.
Other schemes have been proposed to improve performance
and fairness in the presence of unresponsive traffic. Some o
f those schemes
(such as Longest Queue Drop (LQD) and Dynamic Soft Partitioni
ng with Random Drop
(RND) <xref target="SLDC98"/>) were proposed as theoretical f
rameworks and are
typically not available in existing commercial products, whil
e others (such as
Approximate Fairness Through Differential Dropping (AFD) <xre
f target="AFD03" />
have seen some implementation. Advice on the use of
Active Queue Management (AQM) schemes is provided in <xref ta
rget="RFC7567" />.
<xref target="RFC7567" /> recommends self-tuning AQM algorith
ms like those that
the IETF has published in <xref target="RFC8290" />, <xref ta
rget="RFC8033" />,
<xref target="RFC8034" />, and <xref target="RFC9332" />, but
RED is still
appropriate for links with stable bandwidth, if configured ca
refully.</t>
</list></t>
<t>Reactive Versus Preventive Congestion Management Schemes
<list style="symbols">
<t>Reactive (recovery) congestion management policies react
to existing congestion problems. All the policies described
above for
the short and medium timescales can be categorized as being r
eactive.
They are based on monitoring and identifying congestion probl
ems that
exist in the network, and on the initiation of relevant actio
ns to ease
a situation. Reactive congestion management schemes may also
be preventive.</t>
<t>Preventive (predictive/avoidance) policies take proactive
action to prevent congestion based on estimates and predictio
ns of future
congestion problems (e.g., traffic matrix forecasts). Some o
f the policies
described for the long and medium timescales fall into this c
ategory.
Preventive policies do not necessarily respond immediately to
existing
congestion problems. Instead, forecasts of traffic demand an
d workload
distribution are considered, and action may be taken to preve
nt potential
future congestion problems. The schemes described for the sh
ort timescale
can also be used for congestion avoidance because dropping or
marking packets
before queues actually overflow would trigger corresponding r
esponsive traffic sources to
slow down. Preventive congestion management schemes may also
be reactive.</t>
</list></t>
<t>Supply-Side Versus Demand-Side Congestion Management Schemes
<list style="symbols">
<t>Supply-side congestion management policies increase
the effective capacity available to traffic in order to contr
ol or
reduce congestion. This can be accomplished by increasing ca
pacity
or by balancing distribution of traffic over the network. Ca
pacity
planning aims to provide a physical topology and associated l
ink
bandwidths that match or exceed estimated traffic workload an
d traffic
distribution subject to traffic forecasts and budgetary or ot
her
constraints. If the actual traffic distribution does not fit
the
topology derived from capacity planning, then the traffic can
be mapped
onto the topology by using routing control mechanisms, by app
lying path
oriented technologies (e.g., MPLS LSPs and optical channel tr
ails) to
modify the logical topology, or by employing some other load
redistribution
mechanisms.</t>
<t>Demand-side congestion management policies control or
regulate the offered traffic to alleviate congestion problems
. For
example, some of the short timescale mechanisms described ear
lier
as well as policing and rate-shaping mechanisms attempt to re
gulate the
offered load in various ways.</t>
</list></t>
</list></t>
</section>
</section>
<section anchor="IMPCTXT" title="Implementation and Operational Context">
<t>The operational context of Internet TE is
characterized by constant changes that occur at multiple levels of characterized by constant changes that occur at multiple levels of
abstraction. The implementation context demands effective planning, abstraction. The implementation context demands effective planning,
organization, and execution. The planning aspects may involve organization, and execution. The planning aspects may involve
determining prior sets of actions to achieve desired objectives. determining prior sets of actions to achieve desired objectives.
Organizing involves arranging and assigning responsibility to the Organizing involves arranging and assigning responsibility to the
various components of the TE system and coordinating various components of the TE system and coordinating
the activities to accomplish the desired TE objectives. Execution the activities to accomplish the desired TE objectives. Execution
involves measuring and applying corrective or perfective actions to involves measuring and applying corrective or perfective actions to
attain and maintain desired TE goals.</t> attain and maintain desired TE goals.</t>
</section>
</section> </section>
<section anchor="TEPROC" numbered="true" toc="default">
</section> <name>Traffic-Engineering Process Models</name>
<t>This section describes a generic process model that captures the
<section anchor="TEPROC" title="Traffic Engineering Process Models"> high-level practical aspects of Internet traffic engineering in an
operational context. The process model is described as a sequence of
<t>This section describes a generic process model that captures the actions that must be carried out to optimize the performance of an
high-level practical aspects of Internet traffic engineering in an operational network (see also <xref target="RFC2702" format="default"/>
operational context. The process model is described as a sequence and <xref target="AWD2" format="default"/>). This process model may be
of actions that must be carried out to optimize the performance of an enacted explicitly or implicitly, by a software process or by a
operational network (see also <xref target="RFC2702" />, human.</t>
<xref target="AWD2"/>). This process model may be enacted explicitly <t>The TE process model is iterative <xref target="AWD2"
or implicitly, by a software process or by a human.</t> format="default"/>. The four phases of the process model described
below are repeated as a continual sequence:</t>
<t>The TE process model is iterative <xref target="AWD2" />. The four <ol spacing="normal" type="1">
phases of the process model described below are repeated as a continual <li>Define the relevant control policies that govern the operation of
sequence. the network.</li>
<li>Acquire measurement data from the operational network.</li>
<list style="symbols"> <li>Analyze the network state and characterize the traffic workload.
Proactive analysis identifies potential problems that could manifest
<t>Define the relevant control policies that govern the operation in the future. Reactive analysis identifies existing problems and
of the network.</t> determines their causes.</li>
<li>Optimize the performance of the network. This involves a decision
<t>Acquire measurement data from the operational network.</t> process that selects and implements a set of actions from a set of
alternatives given the results of the three previous steps.
<t>Analyze the network state and characterize the traffic workload. Optimization actions may include the use of techniques to control the
Proactive analysis identifies potential problems that could offered traffic and to control the distribution of traffic across the
manifest in the future. Reactive analysis identifies network.</li>
existing problems and determines their causes.</t> </ol>
<section anchor="COMPONENT" numbered="true" toc="default">
<t>Optimize the performance of the network. This involves a decision <name>Components of the Traffic-Engineering Process Model</name>
process which selects and implements a set of actions from a set <t>The key components of the traffic-engineering process model are as
of alternatives given the results of the three previous steps. follows:</t>
Optimization actions may include the use of techniques to control the <ol spacing="normal" type="1"><li>Measurement is crucial to the TE
offered traffic and to control the distribution of traffic across the function. The operational state of a network can only be conclusively
network.</t> determined through measurement. Measurement is also critical to the
optimization function because it provides feedback data that is used
</list></t> by TE control subsystems. This data is used to adaptively optimize
network performance in response to events and stimuli originating
<section anchor="COMPONENT" title="Components of the Traffic Engineering Proce within and outside the network. Measurement in support of the TE
ss Model"> function can occur at different levels of abstraction. For example,
measurement can be used to derive packet-level characteristics, flow-lev
<t>The key components of the traffic engineering process model are as follo el characteristics, user- or customer-level characteristics, traffic-aggregate c
ws. haracteristics, component-level characteristics, and
network-wide characteristics.</li>
<list style="numbers"> <li>Modeling, analysis, and simulation are important aspects of
Internet TE. Modeling involves constructing an abstract or physical
<t>Measurement is crucial to the TE function. The representation that depicts relevant traffic characteristics and
operational state of a network can only be conclusively determined network attributes. A network model is an abstract representation
through measurement. Measurement is also critical to the of the network that captures relevant network features, attributes,
optimization function because it provides feedback data which is and characteristics. Network simulation tools are extremely useful
used by TE control subsystems. This data is for TE. Because of the complexity of realistic quantitative
used to adaptively optimize network performance in response to analysis of network behavior, certain aspects of network performance
events and stimuli originating within and outside the network. studies can only be conducted effectively using simulation.</li>
Measurement in support of the TE function can occur at different <li>Network performance optimization involves resolving network
levels of abstraction. For example, measurement can be used to
derive packet level characteristics, flow level characteristics,
user or customer level characteristics, traffic aggregate
characteristics, component level characteristics, and network-wide
characteristics.</t>
<t>Modeling, analysis, and simulation are important aspects of
Internet TE. Modeling involves constructing an
abstract or physical representation which depicts relevant traffic
characteristics and network attributes. A network model is an
abstract representation of the network which captures relevant
network features, attributes, and characteristics. Network
simulation tools are extremely useful for TE.
Because of the complexity of realistic quantitative analysis of
network behavior, certain aspects of network performance studies
can only be conducted effectively using simulation.</t>
<t>Network performance optimization involves resolving network
issues by transforming such issues into concepts that enable a issues by transforming such issues into concepts that enable a
solution, identification of a solution, and implementation of the solution, identification of a solution, and implementation of the
solution. Network performance optimization can be corrective or solution. Network performance optimization can be corrective or
perfective. In corrective optimization, the goal is to remedy a perfective. In corrective optimization, the goal is to remedy a
problem that has occurred or that is incipient. In perfective problem that has occurred or that is incipient. In perfective
optimization, the goal is to improve network performance even optimization, the goal is to improve network performance even
when explicit problems do not exist and are not anticipated.</t> when explicit problems do not exist and are not anticipated.</li>
</ol>
</list></t> </section>
</section>
</section> <section anchor="TAXI" numbered="true" toc="default">
<name>Taxonomy of Traffic-Engineering Systems</name>
</section> <t>This section presents a short taxonomy of traffic-engineering
<section anchor="TAXI" title="Taxonomy of Traffic Engineering Systems">
<t>This section presents a short taxonomy of traffic engineering
systems constructed based on TE styles and views systems constructed based on TE styles and views
as listed below and described in greater detail in the following as listed below and described in greater detail in the following
subsections of this document.</t> subsections of this document:</t>
<ul spacing="normal">
<t><list style="symbols"> <li><xref target="TIME" format="title"/></li>
<t>Time-dependent versus State-dependent versus Event-dependent</t> <li><xref target="OFFON" format="title"/></li>
<t>Offline versus Online</t> <li><xref target="CENTRAL" format="title"/></li>
<t>Centralized versus Distributed</t> <li><xref target="LOCAL" format="title"/> Information</li>
<t>Local versus Global Information</t> <li><xref target="SCRIPT" format="title"/></li>
<t>Prescriptive versus Descriptive</t> <li><xref target="LOOP" format="title"/></li>
<t>Open Loop versus Closed Loop</t> <li><xref target="TACTIC" format="title"/></li>
<t>Tactical versus Strategic</t> </ul>
</list></t> <section anchor="TIME" numbered="true" toc="default">
<name>Time-Dependent versus State-Dependent versus Event-Dependent</name
<section anchor="TIME" title="Time-Dependent Versus State-Dependent Versus Eve >
nt-Dependent"> <t>Traffic-engineering methodologies can be classified as
time-dependent, state-dependent, or event-dependent. All TE schemes
<t>Traffic engineering methodologies can be classified as time- are considered to be dynamic in this document. Static TE implies that
dependent, state-dependent, or event-dependent. All TE schemes no TE methodology or algorithm is being applied -- it is a feature of
are considered to be dynamic in this document. Static TE implies network planning but lacks the reactive and flexible nature of
that no TE methodology or algorithm is being TE.</t>
applied - it is a feature of network planning, but lacks the <t>In time-dependent TE, historical information based on periodic
reactive and flexible nature of TE.</t> variations in traffic (such as time of day) is used to pre-program
routing and other TE control mechanisms. Additionally, customer
<t>In time-dependent TE, historical information based on periodic subscription or traffic projection may be used. Pre-programmed
variations in traffic (such as time of day) is used to pre-program routing plans typically change on a relatively long timescale (e.g.,
routing and other TE control mechanisms. Additionally, customer daily). Time-dependent algorithms do not attempt to adapt to
subscription or traffic projection may be used. Pre-programmed short-term variations in traffic or changing network conditions. An
routing plans typically change on a relatively long time example of a time-dependent algorithm is a centralized optimizer where
scale (e.g., daily). Time-dependent algorithms do not attempt to the input to the system is a traffic matrix and multi-class QoS
adapt to short-term variations in traffic or changing network conditions. requirements as described in <xref target="MR99" format="default"/>.
An example of a time-dependent algorithm is a centralized Another example of such a methodology is the application of data
optimizer where the input to the system is a traffic matrix and mining to Internet traffic <xref target="AJ19" format="default"/>,
multi-class QoS requirements as described in <xref target="MR99"/>. which enables the use of various machine learning algorithms to
Another example of such a methodology is the application of data mining identify patterns within historically collected datasets about
to Internet traffic <xref target="AJ19"/> which enables the Internet traffic and to extract information in order to guide
use of various machine learning algorithms to identify patterns decision-making and improve efficiency and productivity of
within historically collected datasets about Internet traffic, and to operational processes.</t>
extract information in order to guide decision-making, and to improve <t>State-dependent TE adapts the routing plans based on the current
efficiency and productivity of operational processes.</t> state of the network, which provides additional information on
variations in actual traffic (i.e., perturbations from regular
<t>State-dependent TE adapts the routing plans based on the current variations) that could not be predicted using historical information.
state of the network which provides additional information on Constraint-based routing is an example of state-dependent TE operating
variations in actual traffic (i.e., perturbations from regular in a relatively long timescale. An example of operating in a
variations) that could not be predicted using historical information. relatively short timescale is a load-balancing algorithm described in
Constraint-based routing is an example of state-dependent TE operating <xref target="MATE" format="default"/>. The state of the network can
in a relatively long timescale. An example of operating in a relatively be based on parameters flooded by the routers. Another approach is
short timescale is a load-balancing algorithm described in for a particular router performing adaptive TE to send probe packets
<xref target="MATE"/>. The state of the network can be based on paramete along a path to gather the state of that path. <xref target="RFC6374"
rs format="default"/> defines protocol extensions to collect performance
flooded by the routers. Another approach is for a particular router measurements from MPLS networks. Another approach is for a management
performing adaptive TE to send probe packets along a path to gather the system to gather the relevant information directly from network
state of that path. <xref target="RFC6374" /> defines protocol extension elements using telemetry data collection publication/subscription
s to techniques <xref target="RFC7923" format="default"/>. Timely
collect performance measurements from MPLS networks. Another approach gathering and distribution of state information is critical for
is for a management system to gather the relevant information directly adaptive TE. While time-dependent algorithms are suitable for
from network elements using telemetry data collection "publication/subscr predictable traffic variations, state-dependent algorithms may be
iption" needed to increase network efficiency and to provide resilience to
techniques <xref target="RFC7923" />. Timely gathering and distribution adapt to changes in network state.</t>
of <t>Event-dependent TE methods can also be used for TE path selection.
state information is critical for adaptive TE. While time-dependent algo Event-dependent TE methods are distinct from time-dependent and
rithms state-dependent TE methods in the manner in which paths are selected.
are suitable for predictable traffic variations, state-dependent algorith These algorithms are adaptive and distributed in nature, and they
ms may typically use learning models to find good paths for TE in a network.
be needed to increase network efficiency and to provide resilience to ada While state-dependent TE models typically use available-link-bandwidth
pt to (ALB) flooding <xref target="E.360.1" format="default"/> for TE path
changes in network state.</t> selection, event-dependent TE methods do not require ALB flooding.
Rather, event-dependent TE methods typically search out capacity by
<t>Event-dependent TE methods can also be used for TE path selection. learning models, as in the success-to-the-top (STT) method <xref
Event-dependent TE methods are distinct from time-dependent and target="RFC6601" format="default"/>. ALB flooding can be resource
state-dependent TE methods in the manner in which paths are selected. intensive, since it requires link bandwidth to carry routing protocol
These algorithms are adaptive and distributed in nature, and typically link-state advertisements and processor capacity to process those
use learning models to find good paths for TE in a network. While advertisements; in addition, the overhead of the ALB advertisements and
state-dependent TE models typically use available-link-bandwidth their processing can limit the size of the area and AS. Modeling
(ALB) <xref target="E.360.1" /> flooding for TE path selection, event-dep results suggest that event-dependent TE methods could lead to a
endent TE methods do reduction in ALB flooding overhead without loss of network throughput
not require ALB flooding. Rather, event-dependent TE methods performance <xref target="I-D.ietf-tewg-qos-routing"
typically search out capacity by learning models, as in the format="default"/>.</t>
success-to-the-top (STT) method <xref target="RFC6601" />. ALB flooding <t>A fully functional TE system is likely to use all aspects of
can be resource intensive, time-dependent, state-dependent, and event-dependent methodologies as
since it requires link bandwidth to carry routing protocol link state described in <xref target="HYBRID" format="default"/>.</t>
advertisements, processor capacity to process those advertisements,
and the overhead of the advertisements and their processing can limit
area/Autonomous System (AS) size. Modeling results suggest that
event-dependent TE methods could lead to a reduction in ALB flooding
overhead without loss of network throughput performance
<xref target="I-D.ietf-tewg-qos-routing"/>.</t>
<t>A fully functional TE system is likely to use all aspects of
time-dependent, state-dependent, and event-dependent methodologies as d
escribed in
<xref target="HYBRID" />.</t>
</section>
<section anchor="OFFON" title="Offline Versus Online">
<t>Traffic engineering requires the computation of routing plans. The
computation may be performed offline or online. The computation can
be done offline for scenarios where routing plans need not be
executed in real time. For example, routing plans computed from
forecast information may be computed offline. Typically, offline
computation is also used to perform extensive searches on multi-
dimensional solution spaces.</t>
<t>Online computation is required when the routing plans must adapt to
changing network conditions as in state-dependent algorithms. Unlike
offline computation (which can be computationally demanding), online
computation is geared toward relatively simple and fast calculations to
select routes, fine-tune the allocations of resources, and perform
load balancing.</t>
</section>
<section anchor="CENTRAL" title="Centralized Versus Distributed">
<t>Under centralized control there is a central authority which determines r
outing
plans and perhaps other TE control parameters on behalf of each router.
The
central authority periodically collects network-state information from al
l routers,
and sends routing information to the routers. The update cycle for infor
mation exchange
in both directions is a critical parameter directly impacting the perform
ance of the
network being controlled. Centralized control may need high processing p
ower and high
bandwidth control channels.</t>
<t>Distributed control determines route selection by each router
autonomously based on the router&apos;s view of the state of the network.
The network state information may be obtained by the router using a
probing method or distributed by other routers on a periodic basis
using link state advertisements. Network state information may also
be disseminated under exception conditions. Examples of protocol
extensions used to advertise network link state information are
defined in <xref target="RFC5305"/>, <xref target="RFC6119"/>,
<xref target="RFC7471"/>, <xref target="RFC8570"/>, and
<xref target="RFC8571"/>. See also <xref target="IGPTE" />.</t>
<section anchor="HYBRID" title="Hybrid Systems">
<t>In practice, most TE systems will be a hybrid of central and distribute
d
control. For example, a popular MPLS approach to TE is to use a centra
l
controller based on an active, stateful Path Computation Element (PCE),
but to use routing and signaling
protocols to make local decisions at routers within the network. Local
decisions
may be able to respond more quickly to network events, but may result i
n conflicts
with decisions made by other routers.</t>
<t>Network operations for TE systems may also use a hybrid of offline and
online
computation. TE paths may be precomputed based on stable-state network
information
and planned traffic demands, but may then be modified in the active net
work depending
on variations in network state and traffic load. Furthermore, response
s to network
events may be precomputed offline to allow rapid reactions without furt
her computation,
or may be derived online depending on the nature of the events.</t>
<t>Lastly, note that a fully functional TE system is likely to use all asp
ects of
time-dependent, state-dependent, and event-dependent methodologies as d
escribed in
<xref target="TIME" />.</t>
</section>
<section anchor="SDN" title="Considerations for Software Defined Networking"
>
<t>As discussed in <xref target="ACTN" />, one of the main drivers for
SDN is a decoupling of the network control plane from the data plane
<xref target="RFC7149" />. However, SDN may also combine centralized
control of resources, and facilitate application-to-network interaction
via an application programming interface (API) such as <xref target="RF
C8040" />.
Combining these features provides a flexible network architecture that
can
adapt to network requirements of a variety of higher-layer applications
, a
concept often referred to as the "programmable network" <xref target="R
FC7426" />.</t>
<t>The centralized control aspect of SDN helps improve network resource
utilization compared with distributed network control, where local poli
cy
may often override network-wide optimization goals. In an SDN environm
ent, the
data plane forwards traffic to its desired destination. However, before
traffic reaches the data plane, the logically centralized SDN control p
lane
often determines the path the application traffic will take in
the network. Therefore, the SDN control plane needs to be aware of the
underlying network topology, capabilities and current node and link res
ource
state.</t>
<t>Using a PCE-based SDN control framework <xref target="RFC7491" />, the
available network topology may be discovered by running a passive insta
nce
of OSPF or IS-IS, or via BGP-LS <xref target="RFC7752" />, to generate
a
Traffic Engineering Database (TED, see <xref target="STATE" />).
The PCE is used to compute a path (see
<xref target="PCE" />) based on the TED and available bandwidth, and fu
rther
path optimization may be based on requested objective functions
<xref target="RFC5541" />. When a suitable path has been computed the
programming of the explicit network path may be performed using either
a signaling protocol that traverses the length of the path <xref target
="RFC3209" />
or per-hop with each node being directly programmed <xref target="RFC82
83" /> by the
SDN controller.</t>
<t>By utilizing a centralized approach to network control, additional netw
ork
benefits are also available, including Global Concurrent Optimization (
GCO)
<xref target="RFC5557" />. A GCO path computation request will simulta
neously
use the network topology and a set of new path signaling requests, alon
g with
their respective constraints, for optimal placement in the network.
Correspondingly, a GCO-based computation may be applied to recompute ex
isting
network paths to groom traffic and to mitigate congestion.</t>
</section>
</section>
<section anchor="LOCAL" title="Local Versus Global">
<t>Traffic engineering algorithms may require local and global network-
state information.</t>
<t>Local information is the state of a portion of the domain. Examples inc
lude
the bandwidth and packet loss rate of a particular path, or the state an
d
capabilities of a network link. Local state information may be sufficie
nt
for certain instances of distributed control TE.</t>
<t>Global information is the state of the entire TE domain. Examples inclu
de
a global traffic matrix, and loading information on each link throughout
the
domain of interest. Global state information is typically required with
centralized control. Distributed TE systems may also need global
information in some cases.</t>
</section>
<section anchor="SCRIPT" title="Prescriptive Versus Descriptive">
<t>TE systems may also be classified as prescriptive or descriptive.</t>
<t>Prescriptive traffic engineering evaluates alternatives and
recommends a course of action. Prescriptive TE can
be further categorized as either corrective or perfective.
Corrective TE prescribes a course of action to address an existing or
predicted anomaly. Perfective TE prescribes a course of action to
evolve and improve network performance even when no anomalies are
evident.</t>
<t>Descriptive traffic engineering, on the other hand, characterizes the
state of the network and assesses the impact of various policies
without recommending any particular course of action.</t>
<section anchor="INTENT" title="Intent-Based Networking">
<t>One way to express a service request is through "intent". Intent-Base
d Networking
aims to produce networks that are simpler to manage and operate, requi
ring only
minimal intervention. Intent is defined in <xref target="RFC9315" />
as a set of operational goals (that a network should meet) and outcome
s (that a
network is supposed to deliver), defined in a declarative manner witho
ut specifying
how to achieve or implement them.</t>
<t>Intent provides data and functional abstraction so that users and oper
ators do not
need to be concerned with low-level device configuration or the mechan
isms used to
achieve a given intent. This approach can be conceptually easier for
a user, but may
be less expressive in terms of constraints and guidelines.</t>
<t>Intent-Based Networking is applicable to TE because many of the
high-level objectives may be expressed as "intent." For example, load
balancing,
delivery of services, and robustness against failures. The intent is
converted
by the management system into TE actions within the network.</t>
</section>
</section>
<section anchor="LOOP" title="Open-Loop Versus Closed-Loop">
<t>Open-loop traffic engineering control is where control action does
not use feedback information from the current network state. The
control action may use its own local information for accounting
purposes, however.</t>
<t>Closed-loop traffic engineering control is where control action
utilizes feedback information from the network state. The feedback
information may be in the form of current measurement or recent
historical records.</t>
</section>
<section anchor="TACTIC" title="Tactical versus Strategic">
<t>Tactical traffic engineering aims to address specific performance
problems (such as hot-spots) that occur in the network from a
tactical perspective, without consideration of overall strategic
imperatives. Without proper planning and insights, tactical TE tends
to be ad hoc in nature.</t>
<t>Strategic traffic engineering approaches the TE problem from a more
organized and systematic perspective, taking into consideration the
immediate and longer-term consequences of specific policies and
actions.</t>
</section>
</section>
<section anchor="REVIEW" title="Review of TE Techniques">
<t>This section briefly reviews different TE-related approaches proposed
and implemented in telecommunications and computer networks using
IETF protocols and architectures. These approaches are organized
into three categories:
<list style="symbols">
<t>TE mechanisms which adhere to the definition provided in
<xref target="COMPONENTS" />.</t>
<t>Approaches that rely upon those TE mechanisms.</t>
<t>Techniques that are used by those TE mechanisms and approaches.</t>
</list></t>
<t>The discussion is not intended to be comprehensive. It is primarily
intended to illuminate existing approaches to TE in the Internet. A histo
ric
overview of TE in telecommunications networks was provided in Section 4 of
<xref target="RFC3272" />, and Section 4.6 of that document presented an
outline of some early approaches to TE conducted in other standards bodies
.
It is out of the scope of this document to provide an analysis of the hist
ory
of TE or an inventory of TE-related efforts conducted by other SDOs.</t>
<section anchor="OTHER" title="Overview of IETF Projects Related to Traffic En
gineering">
<t>This subsection reviews a number of IETF activities pertinent to
Internet traffic engineering. Some of these technologies are widely depl
oyed,
others are mature but have seen less deployment, and some are unproven or
still
under development.</t>
<section anchor="TEMech" title="IETF TE Mechanisms">
<section anchor="INTSERV" title="Integrated Services">
<t>The IETF developed the Integrated Services (Intserv) model that requi
res
resources, such as bandwidth and buffers, to be reserved a priori for
a
given traffic flow to ensure that the quality of service requested by
the
traffic flow is satisfied. The Integrated Services model includes ad
ditional
components beyond those used in the best-effort model such as packet
classifiers, packet schedulers, and admission control. A packet clas
sifier
is used to identify flows that are to receive a certain level of serv
ice. A
packet scheduler handles the scheduling of service to different packe
t flows
to ensure that QoS commitments are met. Admission control is used to
determine
whether a router has the necessary resources to accept a new flow.</t
>
<t>The main issue with the Integrated Services model has been scalabilit
y
<xref target="RFC2998"/>, especially in large public IP networks whic
h may
potentially have millions of active traffic flows in transit concurre
ntly.
Pre-Congestion Notification (PCN) <xref target="RFC5559" /> solves th
e scaling
problems of Intserv by using measurement-based admission control (and
flow
termination to handle failures) between edge-nodes. Nodes between th
e edges of
the internetwork have no per-flow operations and the edge nodes can u
se the
Resource Reservation Protocol (RSVP) per-flow or per-aggregate.</t>
<t>A notable feature of the Integrated Services model is that it
requires explicit signaling of QoS requirements from end systems to
routers <xref target="RFC2753"/>. RSVP performs this signaling funct
ion
and is a critical component of the Integrated Services model. RSVP i
s
described in <xref target="RSVP" />.</t>
</section>
<section anchor="DIFFSERV" title="Differentiated Services">
<t>The goal of Differentiated Services (Diffserv) within the IETF was
to devise scalable mechanisms for categorization of traffic into
behavior aggregates, which ultimately allows each behavior aggregate
to be treated differently, especially when there is a shortage of
resources such as link bandwidth and buffer space <xref target="RFC24
75"/>.
One of the primary motivations for Diffserv was to devise alternative
mechanisms for service differentiation in the Internet that mitigate
the scalability issues encountered with the Intserv model.</t>
<t>Diffserv uses the Differentiated Services field in the IP header (the
DS field) consisting of six bits in what was formerly known as the Ty
pe
of Service (TOS) octet. The DS field is used to indicate the forward
ing
treatment that a packet should receive at a transit node <xref target
="RFC2474"/>.
Diffserv includes the concept of Per-Hop Behavior (PHB) groups. Usin
g
the PHBs, several classes of services can be defined using different
classification, policing, shaping, and scheduling rules.</t>
<t>For an end-user of network services to utilize Differentiated
Services provided by its Internet Service Provider (ISP), it may be
necessary for the user to have an SLA with the ISP. An SLA may
explicitly or implicitly specify a Traffic Conditioning Agreement
(TCA) which defines classifier rules as well as metering, marking,
discarding, and shaping rules.</t>
<t>Packets are classified, and possibly policed and shaped at the
ingress to a Diffserv network. When a packet traverses the boundary
between different Diffserv domains, the DS field of the packet may be
re-marked according to existing agreements between the domains.</t>
<t>Differentiated Services allows only a finite number of service
classes to be specified by the DS field. The main advantage of the
Diffserv approach relative to the Intserv model is scalability.
Resources are allocated on a per-class basis and the amount of state
information is proportional to the number of classes rather than to
the number of application flows.</t>
<t>Once the network has been planned and the packets marked at the netwo
rk edge,
the Diffserv model deals with traffic management issues on a per hop
basis. The Diffserv control model consists of a collection of
micro-TE control mechanisms. Other TE capabilities,
such as capacity management (including routing control), are also
required in order to deliver acceptable service quality in Diffserv
networks. The concept of Per Domain Behaviors has been introduced to
better capture the notion of Differentiated Services across a
complete domain <xref target="RFC3086"/>.</t>
<t>Diffserv procedures can also be applied in an MPLS context. See
<xref target="TEDIFFSRV" /> for more information.</t>
</section> </section>
<section anchor="OFFON" numbered="true" toc="default">
<section anchor="SRPolicy" title="Segment Routing Policy"> <name>Offline versus Online</name>
<t>Traffic engineering requires the computation of routing plans. The
<t>SR Policy <xref target="RFC9256" /> is an evolution of Segment Routi computation may be performed offline or online. The computation can
ng (see <xref target="SR" />) be done offline for scenarios where routing plans need not be executed
to enhance the TE capabilities of SR. It is a framework that enable in real time. For example, routing plans computed from forecast
s instantiation of an ordered list of information may be computed offline. Typically, offline computation
segments on a node for implementing a source routing policy with a s is also used to perform extensive searches on multi-dimensional
pecific intent for traffic steering solution spaces.</t>
from that node.</t> <t>Online computation is required when the routing plans must adapt to
changing network conditions as in state-dependent algorithms. Unlike
<t>An SR Policy is identified through the tuple &lt;head-end, color, en offline computation (which can be computationally demanding), online
d-point&gt;. The head-end is the IP address of the computation is geared toward relatively simple and fast calculations
node where the policy is instantiated. The endpoint is the IP addre to select routes, fine-tune the allocations of resources, and perform
ss of the destination of the policy. load balancing.</t>
The color is an index that associates the SR Policy with an intent (
e.g., low latency).</t>
<t>The head-end node is notified of SR Policies and associated SR paths
via configuration or by extensions to protocols
such as PCEP <xref target="RFC8664" /> or BGP <xref target="I-D.ietf
-idr-segment-routing-te-policy" />. Each SR path
consists of a Segment-List (an SR source-routed path), and the head-
end uses the endpoint and color parameters to
classify packets to match the SR policy and so determine along which
path to forward them. If an SR Policy is associated
with a set of SR paths, each is associated with a weight for weighte
d load balancing. Furthermore, multiple SR Policies
may be associated with a set of SR paths to allow multiple traffic f
lows to be placed on the same paths.</t>
<t>An SR Binding SID (BSID) may also be associated with each candidate
path associated with an SR Policy, or with the SR Policy itself. The
head-end node installs a BSID-keyed entry in the forwarding plane an
d assigns it the action of steering packets that match the entry to
the selected path of the SR Policy. This steering can be done in va
rious ways:
<list style="symbols">
<t>SID Steering: Incoming packets have an active SID matching a lo
cal BSID at the head-end.</t>
<t>Per-destination Steering: Incoming packets match a BGP/Service
route which indicates an SR Policy.</t>
<t>Per-flow Steering: Incoming packets match a forwarding array (f
or example, the classic 5-tuple) which indicates an SR Policy.</t>
<t>Policy-based Steering: Incoming packets match a routing policy
which directs them to an SR Policy.</t>
</list></t>
</section> </section>
<section anchor="CENTRAL" numbered="true" toc="default">
<section anchor="QUIC" title="Layer 4 Transport-Based TE"> <name>Centralized versus Distributed</name>
<t>Under centralized control, there is a central authority that
<t>In addition to IP-based TE mechanisms, layer 4 transport-based TE app determines routing plans and perhaps other TE control parameters on
roaches can be considered behalf of each router. The central authority periodically collects
in specific deployment contexts (e.g., data centers, multi-homing). network-state information from all routers and sends routing
For example, the information to the routers. The update cycle for information exchange
3GPP defines the Access Traffic Steering, Switching, and Splitting (A in both directions is a critical parameter directly impacting the
TSSS) <xref target="ATSSS" /> performance of the network being controlled. Centralized control may
service functions as follows. need high processing power and high bandwidth control channels.</t>
<list style="hanging"> <t>Distributed control determines route selection by each router
<t hangText='Access Traffic Steering:'> This is the selection of a autonomously based on the router's view of the state of the network.
n access network for a The network state information may be obtained by the router using a
new flow and the transfer of the traffic of that flow over the probing method or distributed by other routers on a periodic basis
selected access network.</t> using link-state advertisements. Network state information may also
be disseminated under exception conditions. Examples of protocol
<t hangText='Access Traffic Switching:'> This is the migration of extensions used to advertise network link-state information are
all packets of an defined in <xref target="RFC5305" format="default"/>, <xref
ongoing flow from one access network to another access network. target="RFC6119" format="default"/>, <xref target="RFC7471"
Only one access format="default"/>, <xref target="RFC8570" format="default"/>, and
network is in use at a time.</t> <xref target="RFC8571" format="default"/>. See also <xref
target="IGPTE" format="default"/>.</t>
<t hangText='Access Traffic Splitting:'> This is about forwarding <section anchor="HYBRID" numbered="true" toc="default">
the packets of a flow <name>Hybrid Systems</name>
across multiple access networks simultaneously.</t> <t>In practice, most TE systems will be a hybrid of central and
</list></t> distributed control. For example, a popular MPLS approach to TE is
to use a central controller based on an active, stateful Path
<t>The control plane is used to provide hosts and specific network devic Computation Element (PCE) but to use routing and signaling protocols
es with a set of policies to make local decisions at routers within the network. Local
that specify which flows are eligible to use the ATSSS service. The decisions may be able to respond more quickly to network events but
traffic that matches an ATSSS may result in conflicts with decisions made by other routers.</t>
policy can be distributed among the available access networks followi <t>Network operations for TE systems may also use a hybrid of
ng one of the following four modes. offline and online computation. TE paths may be precomputed based
<list style="hanging"> on stable-state network information and planned traffic demands but
<t hangText='Active-Standby:'> The traffic is forwarded via a spec may then be modified in the active network depending on variations
ific access (called "active in network state and traffic load. Furthermore, responses to
access") and switched to another access (called "standby access network events may be precomputed offline to allow rapid reactions
") when the active access is unavailable.</t> without further computation or may be derived online depending on
the nature of the events.</t>
<t hangText='Priority-based:'> Network accesses are assigned prior </section>
ity levels that indicate which <section anchor="SDN" numbered="true" toc="default">
network access is to be used first. The traffic associated wit <name>Considerations for Software-Defined Networking</name>
h the matching flow will be <t>As discussed in <xref target="ACTN" format="default"/>, one of
steered onto the network access with the highest priority until the main drivers for Software-Defined Networking (SDN) is a
congestion is detected, then the decoupling of the network control plane from the data plane <xref
overflow will be forwarded over the next highest priority acces target="RFC7149" format="default"/>. However, SDN may also combine
s.</t> centralized control of resources and facilitate
application-to-network interaction via an Application Programming
<t hangText='Load-Balancing:'> The traffic is distributed among th Interface (API), such as the one described in <xref target="RFC8040"
e available access networks following format="default"/>. Combining these features provides a flexible
a distribution ratio (e.g., 75% - 25%).</t> network architecture that can adapt to the network requirements of a
variety of higher-layer applications, a concept often referred to as
<t hangText='Smallest Delay:'> The traffic is forwarded via the ac the "programmable network" <xref target="RFC7426"
cess that presents the smallest format="default"/>.</t>
round-trip-time (RTT).</t> <t>The centralized control aspect of SDN helps improve network
</list></t> resource utilization compared with distributed network control,
where local policy may often override network-wide optimization
<t>For resource management purposes, hosts and network devices support m goals. In an SDN environment, the data plane forwards traffic to
eans such as congestion control, its desired destination. However, before traffic reaches the data
RTT measurement, and packet scheduling.</t> plane, the logically centralized SDN control plane often determines
the path the application traffic will take in the network.
<t>For TCP traffic, Multipath TCP <xref target="RFC8684" /> and the 0-RT Therefore, the SDN control plane needs to be aware of the underlying
T Convert Protocol <xref target="RFC8803" /> network topology, capabilities, and current node and link resource
are used to provide the ATSSS service.</t> state.</t>
<t>Using a PCE-based SDN control framework <xref target="RFC7491"
<t>Multipath QUIC <xref target="I-D.ietf-quic-multipath" /> and Proxying format="default"/>, the available network topology may be discovered
UDP in HTTP <xref target="RFC9298" /> by running a passive instance of OSPF or IS-IS, or via BGP Link
are used to provide the ATSSS service for UDP traffic. Note that QUI State (BGP-LS) <xref target="RFC9552" format="default"/>), to
C <xref target="RFC9000" /> natively generate a Traffic Engineering Database (TED) (see <xref
support the switching and steering functions. Indeed, QUIC supports target="STATE" format="default"/>). The PCE is used to compute a
a connection migration procedure path (see <xref target="PCE" format="default"/>) based on the TED
that allows peers to change their layer 4 transport coordinates (IP a and available bandwidth, and further path optimization may be based
ddresses, port numbers) without breaking on requested objective functions <xref target="RFC5541"
the underlying QUIC connection.</t> format="default"/>. When a suitable path has been computed, the
programming of the explicit network path may be either performed
<t>Extensions to the Datagram Congestion Control Protocol (MP-DCCP) <xre using a signaling protocol that traverses the length of the path
f target="RFC4340" /> to support multipath <xref target="RFC3209" format="default"/> or performed per-hop with
operations are defined in <xref target="I-D.ietf-tsvwg-multipath-dccp each node being directly programmed <xref target="RFC8283"
" />.</t> format="default"/> by the SDN controller.</t>
<t>By utilizing a centralized approach to network control,
additional network benefits are also available, including Global
Concurrent Optimization (GCO) <xref target="RFC5557"
format="default"/>. A GCO path computation request will
simultaneously use the network topology and a set of new path
signaling requests, along with their respective constraints, for
optimal placement in the network. Correspondingly, a GCO-based
computation may be applied to recompute existing network paths to
groom traffic and to mitigate congestion.</t>
</section>
</section> </section>
<section anchor="LOCAL" numbered="true" toc="default">
<section anchor="DETNET" title="Deterministic Networking"> <name>Local versus Global</name>
<t>Traffic-engineering algorithms may require local and global
<t>Deterministic Networking (DetNet) <xref target="RFC8655" /> is an ar network-state information.</t>
chitecture for applications with <t>Local information is the state of a portion of the domain.
critical timing and reliability requirements. The layered architect Examples include the bandwidth and packet loss rate of a particular
ure particularly focuses on path or the state and capabilities of a network link. Local state
developing DetNet service capabilities in the data plane <xref targe information may be sufficient for certain instances of distributed
t="RFC8938" />. control TE.</t>
The DetNet service sub-layer provides a set of Packet Replication, E <t>Global information is the state of the entire TE domain. Examples
limination, and Ordering Functions (PREOF) include a global traffic matrix and loading information on each link
to provide end-to-end service assurance. The DetNet forwarding sub- throughout the domain of interest. Global state information is
layer provides corresponding typically required with centralized control. Distributed TE systems
forwarding assurance (low packet loss, bounded latency, and in-order may also need global information in some cases.</t>
delivery) functions using resource
allocations and explicit route mechanisms.</t>
<t>The separation into two sub-layers allows a greater flexibility to a
dapt DetNet capability over a number of TE
data plane mechanisms such as IP, MPLS, and Segment Routing. More i
mportantly it interconnects IEEE 802.1 Time
Sensitive Networking (TSN) <xref target="RFC9023" /> deployed in Ind
ustry Control and Automation
Systems (ICAS).</t>
<t>DetNet can be seen as a specialized branch of TE, since it sets up e
xplicit optimized paths with allocation of
resources as requested. A DetNet application can express its QoS at
tributes or traffic behavior using any
combination of DetNet functions described in sub-layers. They are t
hen distributed and provisioned using well-
established control and provisioning mechanisms adopted for traffic
engineering.</t>
<t>In DetNet, a considerable amount of state information is required to
maintain per-flow queuing disciplines and resource
reservation for a large number of individual flows. This can be qui
te challenging for network operations during
network events such as faults, change in traffic volume or re-provis
ioning. Therefore, DetNet recommends support
for aggregated flows, however, it still requires a large amount of c
ontrol signaling to establish and maintain
DetNet flows.</t>
<t>Note that DetNet might suffer from some of the scalability concerns
described for Intserv in <xref target="INTSERV" />,
but the scope of DetNet&apos;s deployment scenarios is smaller and s
o less exposed to scaling issues.</t>
</section> </section>
<section anchor="SCRIPT" numbered="true" toc="default">
</section> <name>Prescriptive versus Descriptive</name>
<t>TE systems may also be classified as prescriptive or descriptive.</t>
<section anchor="TEapproach" title="IETF Approaches Relying on TE Mechanisms <t>Prescriptive traffic engineering evaluates alternatives and
"> recommends a course of action. Prescriptive TE can be further
categorized as either corrective or perfective. Corrective TE
<section anchor="ALTO" title="Application-Layer Traffic Optimization"> prescribes a course of action to address an existing or predicted
anomaly. Perfective TE prescribes a course of action to evolve and
<t>This document describes various TE mechanisms available in the netwo improve network performance even when no anomalies are evident.</t>
rk. However, distributed <t>Descriptive traffic engineering, on the other hand, characterizes
applications in general and, in particular, bandwidth-greedy P2P app the state of the network and assesses the impact of various policies
lications that are used, without recommending any particular course of action.</t>
for example, for file sharing, cannot directly use those techniques. <section anchor="INTENT" numbered="true" toc="default">
As per <xref target="RFC5693" />, <name>Intent-Based Networking</name>
applications could greatly improve traffic distribution and quality <t>One way to express a service request is through "intent".
by cooperating with external Intent-Based Networking aims to produce networks that are simpler to
services that are aware of the network topology. Addressing the App manage and operate, requiring only minimal intervention. Intent is
lication-Layer Traffic defined in <xref target="RFC9315" format="default"/> as follows:</t>
Optimization (ALTO) problem means, on the one hand, deploying an ALT <blockquote>
O service to provide A set of operational goals (that a network should meet) and outcomes
applications with information regarding the underlying network (e.g. (that a network is supposed to deliver) defined in a declarative
, basic network location manner without specifying how to achieve or implement
structure and preferences of network paths) and, on the other hand, them.</blockquote>
enhancing applications in <t>Intent provides data and functional abstraction so that users and
order to use such information to perform better-than-random selectio operators do not need to be concerned with low-level device
n of the endpoints with configuration or the mechanisms used to achieve a given intent.
which they establish connections.</t> This approach can be conceptually easier for a user but may be less
expressive in terms of constraints and guidelines.</t>
<t>The basic function of ALTO is based on abstract maps of a network. <t>Intent-Based Networking is applicable to TE because many of the
These maps provide a high-level objectives may be expressed as intent (for example,
simplified view, yet enough information about a network for applicat load balancing, delivery of services, and robustness against
ions to effectively utilize failures). The intent is converted by the management system into TE
them. Additional services are built on top of the maps. <xref targ actions within the network.</t>
et="RFC7285" /> describes a </section>
protocol implementing the ALTO services as an information-publishing
interface that allows a
network to publish its network information to network applications.
This information can include
network node locations, groups of node-to-node connectivity arranged
by cost according to
configurable granularities, and end-host properties. The informatio
n
published by the ALTO Protocol should benefit both the network and t
he applications. The ALTO
Protocol uses a REST-ful design and encodes its requests and respons
es using JSON
<xref target="RFC8259" /> with a modular design by dividing ALTO inf
ormation publication into
multiple ALTO services (e.g., the Map service, the Map-Filtering Ser
vice, the Endpoint Property
Service, and the Endpoint Cost Service).</t>
<t><xref target="RFC8189" /> defines a new service that allows an ALTO
Client to retrieve several
cost metrics in a single request for an ALTO filtered cost map and e
ndpoint cost map.
<xref target="RFC8896" /> extends the ALTO cost information service
so that applications decide
not only 'where' to connect, but also 'when'. This is useful for ap
plications that need to perform
bulk data transfer and would like to schedule these transfers during
an off-peak hour, for example.
<xref target="RFC9439" /> introduces network performance metrics, in
cluding
network delay, jitter, packet loss rate, hop count, and bandwidth.
The ALTO server may derive and
aggregate such performance metrics from BGP-LS (see <xref target="BG
PLS" />) or IGP-TE (see
<xref target="IGPTE" />), or management tools, and then expose the i
nformation to allow applications
to determine 'where' to connect based on network performance criteri
a. The ALTO WG is evaluating the use
of network TE properties while making application decisions for new
use cases such as Edge computing
and Datacenter interconnect.</t>
</section> </section>
<section anchor="LOOP" numbered="true" toc="default">
<section anchor="ACTN" title="Network Virtualization and Abstraction"> <name>Open-Loop versus Closed-Loop</name>
<t>Open-loop traffic-engineering control is where control action does
<t>One of the main drivers for Software Defined Networking (SDN) not use feedback information from the current network state. However,
<xref target="RFC7149" /> is a decoupling of the network control the control action may use its own local information for accounting
plane from the data plane. This separation has been achieved for purposes.</t>
TE networks with the development of MPLS/GMPLS (see <xref target="MP <t>Closed-loop traffic-engineering control is where control action
LS" /> utilizes feedback information from the network state. The feedback
and <xref target="GMPLS" />) and the PCE (<xref target="PCE" />). O information may be in the form of current measurement or recent
ne of the historical records.</t>
advantages of SDN is its logically centralized control regime that a
llows a
full view of the underlying networks. Centralized control in SDN he
lps
improve network resource utilization compared with distributed netwo
rk control.</t>
<t>Abstraction and Control of TE Networks (ACTN) <xref target="RFC8453"
/>
defines a hierarchical SDN architecture which describes the function
al
entities and methods for the coordination of resources across multip
le
domains, to provide composite traffic-engineered services. ACTN
facilitates composed, multi-domain connections and provides them to
the user. ACTN
is focused on:
<list style="symbols">
<t>Abstraction of the underlying network resources and how they ar
e
provided to higher-layer applications and customers.</t>
<t>Virtualization of underlying resources for use by the customer,
application, or service. The creation of a virtualized environ
ment
allows operators to view and control multi-domain networks as a
single
virtualized network.</t>
<t>Presentation to customers of networks as a virtual network via
open and programmable interfaces.</t>
</list></t>
<t>The ACTN managed infrastructure is built from traffic-engineered net
work
resources, which may include statistical packet bandwidth, physical
forwarding plane sources (such as wavelengths and time slots), forwa
rding
and cross-connect capabilities. The type of network virtualization
seen in
ACTN allows customers and applications (tenants) to utilize and inde
pendently
control allocated virtual network resources as if they were
physically their own resource. The ACTN network is "sliced", with t
enants
being given a different partial and abstracted topology view of the
physical
underlying network.</t>
</section> </section>
<section anchor="TACTIC" numbered="true" toc="default">
<section anchor="SLICE" title="Network Slicing"> <name>Tactical versus Strategic</name>
<t>Tactical traffic engineering aims to address specific performance
<t>An IETF Network Slice is a logical network topology connecting a num problems (such as hotspots) that occur in the network from a
ber of tactical perspective, without consideration of overall strategic
endpoints using a set of shared or dedicated network resources imperatives. Without proper planning and insights, tactical TE tends
<xref target="I-D.ietf-teas-ietf-network-slices" />. The resources to be ad hoc in nature.</t>
are used <t>Strategic traffic-engineering approaches the TE problem from a more
to satisfy specific Service Level Objectives (SLOs) specified by the organized and systematic perspective, taking into consideration the
consumer.</t> immediate and longer-term consequences of specific policies and
actions.</t>
<t>IETF network slices are not, of themselves, TE constructs. However,
a network operator that offers
IETF network slices is likely to use many TE tools in order to manag
e their network and provide the
services.</t>
<t>IETF Network Slices are defined such that they are independent of th
e underlying infrastructure
connectivity and technologies used. From a customer&apos;s perspect
ive, an IETF Network Slice looks
like a VPN connectivity matrix with additional information about the
level of service that the
customer requires between the endpoints. From an operator&apos;s pe
rspective, the IETF Network Slice
looks like a set of routing or tunneling instructions with the netwo
rk resource reservations necessary
to provide the required service levels as specified by the SLOs. Th
e concept of an IETF network slice
is consistent with an enhanced VPN (VPN+) <xref target="I-D.ietf-tea
s-enhanced-vpn" />.</t>
</section> </section>
</section> </section>
<section anchor="REVIEW" numbered="true" toc="default">
<name>Review of TE Techniques</name>
<t>This section briefly reviews different TE-related approaches proposed
and implemented in telecommunications and computer networks using IETF
protocols and architectures. These approaches are organized into three
categories:</t>
<ul spacing="normal">
<li>TE mechanisms that adhere to the definition provided in <xref
target="COMPONENTS" format="default"/></li>
<li>Approaches that rely upon those TE mechanisms</li>
<li>Techniques that are used by those TE mechanisms and
approaches</li>
</ul>
<t>The discussion is not intended to be comprehensive. It is primarily
intended to illuminate existing approaches to TE in the Internet. A
historic overview of TE in telecommunications networks was provided in
<xref target="RFC3272" sectionFormat="of" section="4"/>, and Section
<xref target="RFC3272" sectionFormat="bare" section="4.6"/> of that
document presented an outline of some early approaches to TE conducted
in other standards bodies. It is out of the scope of this document to
provide an analysis of the history of TE or an inventory of TE-related
efforts conducted by other Standards Development Organizations
(SDOs).</t>
<section anchor="OTHER" numbered="true" toc="default">
<name>Overview of IETF Projects Related to Traffic Engineering</name>
<t>This subsection reviews a number of IETF activities pertinent to
Internet traffic engineering. Some of these technologies are widely
deployed, others are mature but have seen less
deployment, and some are unproven or are still under
development.</t>
<section anchor="TEMech" numbered="true" toc="default">
<name>IETF TE Mechanisms</name>
<section anchor="INTSERV" numbered="true" toc="default">
<name>Integrated Services</name>
<t>The IETF developed the Integrated Services (Intserv) model that
requires resources, such as bandwidth and buffers, to be reserved
a priori for a given traffic flow to ensure that the QoS requested
by the traffic flow is satisfied. The Intserv model includes
additional components beyond those used in the best-effort model
such as packet classifiers, packet schedulers, and admission
control. A packet classifier is used to identify flows that are
to receive a certain level of service. A packet scheduler handles
the scheduling of service to different packet flows to ensure that
QoS commitments are met. Admission control is used to determine
whether a router has the necessary resources to accept a new
flow.</t>
<t>The main issue with the Intserv model has been scalability
<xref target="RFC2998" format="default"/>, especially in large
public IP networks that may potentially have millions of active
traffic flows in transit concurrently. Pre-Congestion
Notification (PCN) <xref target="RFC5559" format="default"/>
solves the scaling problems of Intserv by using measurement-based
admission control (and flow termination to handle failures)
between edge nodes. Nodes between the edges of the internetwork
have no per-flow operations, and the edge nodes can use the
Resource Reservation Protocol (RSVP) per-flow or
per-aggregate.</t>
<t>A notable feature of the Intserv model is that it requires
explicit signaling of QoS requirements from end systems to routers
<xref target="RFC2753" format="default"/>. RSVP performs this
signaling function and is a critical component of the Intserv
model. RSVP is described in <xref target="RSVP"
format="default"/>.</t>
</section>
<section anchor="DIFFSERV" numbered="true" toc="default">
<name>Differentiated Services</name>
<t>The goal of Differentiated Services (Diffserv) within the IETF
was to devise scalable mechanisms for categorization of traffic
into behavior aggregates, which ultimately allows each behavior
aggregate to be treated differently, especially when there is a
shortage of resources, such as link bandwidth and buffer space
<xref target="RFC2475" format="default"/>. One of the primary
motivations for Diffserv was to devise alternative mechanisms for
service differentiation in the Internet that mitigate the
scalability issues encountered with the Intserv model.</t>
<t>Diffserv uses the Differentiated Services field in the IP
header (the DS field) consisting of six bits in what was formerly
known as the Type of Service (TOS) octet. The DS field is used to
indicate the forwarding treatment that a packet should receive at
a transit node <xref target="RFC2474" format="default"/>.
Diffserv includes the concept of Per-Hop Behavior (PHB) groups.
Using the PHBs, several classes of services can be defined using
different classification, policing, shaping, and scheduling
rules.</t>
<t>For an end user of network services to utilize Diffserv
provided by its Internet Service Provider (ISP), it may
be necessary for the user to have an SLA with the ISP. An SLA may
explicitly or implicitly specify a Traffic Conditioning Agreement
(TCA) that defines classifier rules as well as metering, marking,
discarding, and shaping rules.</t>
<t>Packets are classified and possibly policed and shaped at the
ingress to a Diffserv network. When a packet traverses the
boundary between different Diffserv domains, the DS field of the
packet may be re-marked according to existing agreements between
the domains.</t>
<t>Diffserv allows only a finite number of service
classes to be specified by the DS field. The main advantage of
the Diffserv approach relative to the Intserv model is
scalability. Resources are allocated on a per-class basis, and the
amount of state information is proportional to the number of
classes rather than to the number of application flows.</t>
<t>Once the network has been planned and the packets have been
marked at the network edge, the Diffserv model deals with traffic
management issues on a per-hop basis. The Diffserv control model
consists of a collection of micro-TE control mechanisms. Other TE
capabilities, such as capacity management (including routing
control), are also required in order to deliver acceptable service
quality in Diffserv networks. The concept of "Per-Domain
Behaviors" has been introduced to better capture the notion of
Diffserv across a complete domain <xref target="RFC3086"
format="default"/>.</t>
<t>Diffserv procedures can also be applied in an MPLS context. See
<xref target="TEDIFFSRV" format="default"/> for more information.</t>
</section>
<section anchor="SRPolicy" numbered="true" toc="default">
<name>SR Policy</name>
<t>SR Policy <xref target="RFC9256" format="default"/> is an
evolution of SR (see <xref target="SR"
format="default"/>) to enhance the TE capabilities of SR. It is a
framework that enables instantiation of an ordered list of
segments on a node for implementing a source routing policy with a
specific intent for traffic steering from that node.</t>
<t>An SR Policy is identified through the tuple &lt;headend,
color, endpoint&gt;. The headend is the IP address of the node
where the policy is instantiated. The endpoint is the IP address
of the destination of the policy. The color is an index that
associates the SR Policy with an intent (e.g., low latency).</t>
<t>The headend node is notified of SR Policies and associated SR
paths via configuration or by extensions to protocols such as the Pa
th Computation Element Communication Protocol (PCEP)
<xref target="RFC8664" format="default"/> or BGP <xref
target="I-D.ietf-idr-segment-routing-te-policy"
format="default"/>. Each SR path consists of a segment list (an
SR source-routed path), and the headend uses the endpoint and
color parameters to classify packets to match the SR Policy and so
determine along which path to forward them. If an SR Policy is
associated with a set of SR paths, each is associated with a
weight for weighted load balancing. Furthermore, multiple SR
Policies may be associated with a set of SR paths to allow
multiple traffic flows to be placed on the same paths.</t>
<t>An SR Binding SID (BSID) may also be associated with each
candidate path associated with an SR Policy or
with the SR Policy itself. The headend node installs a
BSID-keyed entry in the forwarding plane and assigns it the action
of steering packets that match the entry to the selected path of
the SR Policy. This steering can be done in various ways:</t>
<dl newline="false" spacing="normal">
<dt>SID Steering:</dt><dd>Incoming packets have an active Segment
Identifier (SID) matching a local BSID at
the headend.</dd>
<dt>Per-destination Steering:</dt><dd>
Incoming packets match a BGP/Service route, which indicates
an SR Policy.</dd>
<dt>Per-flow Steering:</dt><dd>
Incoming packets match a forwarding array (for example, the
classic 5-tuple), which indicates an SR Policy.</dd>
<dt>Policy-based Steering:</dt><dd>
Incoming packets match a routing policy, which directs them
to an SR Policy.</dd>
</dl>
</section>
<section anchor="QUIC" numbered="true" toc="default">
<name>Layer 4 Transport-Based TE</name>
<t>In addition to IP-based TE mechanisms, Layer 4 transport-based
TE approaches can be considered in specific deployment contexts
(e.g., data centers and multi-homing). For example, the 3GPP define
s
the Access Traffic Steering, Switching, and Splitting (ATSSS)
<xref target="ATSSS" format="default"/> service functions as
follows:
</t>
<dl newline="false" spacing="normal">
<dt>Access Traffic Steering:</dt>
<dd>This is the selection of an access network for a new flow
and the transfer of the traffic of that flow over the selected
access network.</dd>
<dt>Access Traffic Switching:</dt>
<dd>This is the migration of all packets of an ongoing flow from
one access network to another access network. Only one access
network is in use at a time.</dd>
<dt>Access Traffic Splitting:</dt>
<dd>This is about forwarding the packets of a flow across
multiple access networks simultaneously.</dd>
</dl>
<t>The control plane is used to provide hosts and specific network
devices with a set of policies that specify which flows are
eligible to use the ATSSS service. The traffic that matches an
ATSSS policy can be distributed among the available access
networks following one of the following four modes:</t>
<dl newline="false" spacing="normal">
<dt>Active-Standby:</dt>
<dd>The traffic is forwarded via a specific access (called
"active access") and switched to another access (called "standby
access") when the active access is unavailable.</dd>
<dt>Priority-based:</dt>
<dd>Network accesses are assigned priority levels that indicate
which network access is to be used first. The traffic
associated with the matching flow will be steered onto the
network access with the highest priority until congestion is
detected. Then, the overflow will be forwarded over the next
highest priority access.</dd>
<dt>Load-Balancing:</dt>
<dd>The traffic is distributed among the available access
networks following a distribution ratio (e.g., 75% to 25%).</dd>
<dt>Smallest Delay:</dt>
<dd>The traffic is forwarded via the access that presents the
smallest round-trip time (RTT).</dd>
</dl>
<t>For resource management purposes, hosts and network devices
support means such as congestion control, RTT measurement, and
packet scheduling.</t>
<t>For TCP traffic, Multipath TCP <xref target="RFC8684"
format="default"/> and the 0-RTT Convert Protocol <xref
target="RFC8803" format="default"/> are used to provide the ATSSS
service.</t>
<t>Multipath QUIC <xref target="I-D.ietf-quic-multipath"
format="default"/> and Proxying UDP in HTTP
<xref target="RFC9298" format="default"/> are used to provide the
ATSSS service for UDP traffic. Note that QUIC <xref
target="RFC9000" format="default"/> supports the
switching and steering functions. Indeed, QUIC supports a
connection migration procedure that allows peers to change their
Layer 4 transport coordinates (IP addresses, port numbers) without
breaking the underlying QUIC connection.</t>
<t>Extensions to the Datagram Congestion Control Protocol
(DCCP) <xref target="RFC4340" format="default"/> to support
multipath operations are defined in <xref
target="I-D.ietf-tsvwg-multipath-dccp" format="default"/>.</t>
</section>
<section anchor="DETNET" numbered="true" toc="default">
<name>Deterministic Networking</name>
<t>Deterministic Networking (DetNet) <xref target="RFC8655" format="
default"/> is an
architecture for applications with critical timing and reliability
requirements. The layered architecture particularly focuses on
developing DetNet service capabilities in the data plane <xref
target="RFC8938" format="default"/>. The DetNet service sub-layer
provides a set of Packet Replication, Elimination, and Ordering
Functions (PREOF) to provide end-to-end service assurance. The
DetNet forwarding sub-layer provides corresponding forwarding
assurance (low packet loss, bounded latency, and in-order
delivery) functions using resource allocations and explicit route
mechanisms.</t>
<t>The separation into two sub-layers allows a greater flexibility
to adapt DetNet capability over a number of TE data plane
mechanisms, such as IP, MPLS, and SR. More
importantly, it interconnects IEEE 802.1 Time Sensitive Networking
(TSN) <xref target="RFC9023" format="default"/> deployed in
Industry Control and Automation Systems (ICAS).</t>
<t>DetNet can be seen as a specialized branch of TE, since it sets
up explicit optimized paths with allocation of resources as
requested. A DetNet application can express its QoS attributes or
traffic behavior using any combination of DetNet functions
described in sub-layers. They are then distributed and
provisioned using well-established control and provisioning
mechanisms adopted for traffic engineering.</t>
<t>In DetNet, a considerable amount of state information is
required to maintain per-flow queuing disciplines and resource
reservation for a large number of individual flows. This can be
quite challenging for network operations during network events,
such as faults, change in traffic volume, or reprovisioning.
Therefore, DetNet recommends support for aggregated flows;
however, it still requires a large amount of control signaling to
establish and maintain DetNet flows.</t>
<t>Note that DetNet might suffer from some of the scalability
concerns described for Intserv in <xref target="INTSERV"
format="default"/>, but the scope of DetNet's deployment scenarios
is smaller and therefore less exposed to scaling issues.</t>
</section>
</section>
<section anchor="TEapproach" numbered="true" toc="default">
<name>IETF Approaches Relying on TE Mechanisms</name>
<section anchor="ALTO" numbered="true" toc="default">
<name>Application-Layer Traffic Optimization</name>
<t>This document describes various TE mechanisms available in the
network. However, in general, distributed applications
(particularly, bandwidth-greedy P2P applications that are used for
file sharing, for example) cannot directly use those techniques.
As per <xref target="RFC5693" format="default"/>, applications
could greatly improve traffic distribution and quality by
cooperating with external services that are aware of the network
topology. Addressing the Application-Layer Traffic Optimization
(ALTO) problem means, on the one hand, deploying an ALTO service
to provide applications with information regarding the underlying
network (e.g., basic network location structure and preferences of
network paths) and, on the other hand, enhancing applications in
order to use such information to perform better-than-random
selection of the endpoints with which they establish
connections.</t>
<section anchor="TEtech" title="IETF Techniques Used by TE Mechanisms"> <t>The basic function of ALTO is based on abstract maps of a
network. These maps provide a simplified view, yet enough
<section anchor="CSPF" title="Constraint-Based Routing"> information about a network for applications to effectively
utilize them. Additional services are built on top of the maps.
<t>Constraint-based routing refers to a class of routing systems that <xref target="RFC7285" format="default"/> describes a protocol
implementing the ALTO services as an information-publishing
interface that allows a network to publish its network information
to network applications. This information can include network
node locations, groups of node-to-node connectivity arranged by
cost according to configurable granularities, and end-host
properties. The information published by the ALTO Protocol should
benefit both the network and the applications. The ALTO Protocol
uses a REST-ful design and encodes its requests and responses
using JSON <xref target="RFC8259" format="default"/> with a
modular design by dividing ALTO information publication into
multiple ALTO services (e.g., the Map Service, the Map-Filtering
Service, the Endpoint Property Service, and the Endpoint Cost
Service).</t>
<t><xref target="RFC8189" format="default"/> defines a new service
that allows an ALTO Client to retrieve several cost metrics in a
single request for an ALTO filtered cost map and endpoint cost
map. <xref target="RFC8896" format="default"/> extends the ALTO
cost information service so that applications decide not only
"where" to connect but also "when". This is useful for
applications that need to perform bulk data transfer and would
like to schedule these transfers during an off-peak hour, for
example. <xref target="RFC9439" format="default"/> introduces
network performance metrics, including network delay, jitter,
packet loss rate, hop count, and bandwidth. The ALTO server may
derive and aggregate such performance metrics from BGP-LS (see
<xref target="BGPLS" format="default"/>), IGP-TE (see <xref
target="IGPTE" format="default"/>), or management tools and then
expose the information to allow applications to determine "where"
to connect based on network performance criteria. The ALTO
Working Group is evaluating the use of network TE properties while
making application decisions for new use cases such as edge
computing and data-center interconnect.</t>
</section>
<section anchor="ACTN" numbered="true" toc="default">
<name>Network Virtualization and Abstraction</name>
<t>One of the main drivers for SDN
<xref target="RFC7149" format="default"/> is a decoupling of the
network control plane from the data plane. This separation has
been achieved for TE networks with the development of MPLS and GMPLS
(see Sections <xref target="MPLS" format="counter"/> and <xref
target="GMPLS" format="counter"/>, respectively) and the PCE (see <x
ref target="PCE"
format="default"/>). One of the advantages of SDN is its
logically centralized control regime that allows a full view of
the underlying networks. Centralized control in SDN helps improve
network resource utilization compared with distributed network
control.</t>
<t>Abstraction and Control of TE Networks (ACTN) <xref
target="RFC8453" format="default"/> defines a hierarchical SDN
architecture that describes the functional entities and methods
for the coordination of resources across multiple domains, to
provide composite traffic-engineered services. ACTN facilitates
composed, multi-domain connections and provides them to the user.
ACTN is focused on:
</t>
<ul spacing="normal">
<li>Abstraction of the underlying network resources and how they
are provided to higher-layer applications and customers.</li>
<li>Virtualization of underlying resources for use by the
customer, application, or service. The creation of a
virtualized environment allows operators to view and control
multi-domain networks as a single virtualized network.</li>
<li>Presentation to customers of networks as a virtual network
via open and programmable interfaces.</li>
</ul>
<t>The ACTN managed infrastructure is built from
traffic-engineered network resources, which may include
statistical packet bandwidth, physical forwarding-plane sources
(such as wavelengths and time slots), and forwarding and cross-conne
ct
capabilities. The type of network virtualization seen in ACTN
allows customers and applications (tenants) to utilize and
independently control allocated virtual network resources as if
they were physically their own resource. The ACTN network is
sliced, with tenants being given a different partial and
abstracted topology view of the physical underlying network.</t>
</section>
<section anchor="SLICE" numbered="true" toc="default">
<name>Network Slicing</name>
<t>An IETF Network Slice is a logical network topology connecting
a number of endpoints using a set of shared or dedicated network
resources <xref target="I-D.ietf-teas-ietf-network-slices"
format="default"/>. The resources are used to satisfy specific
SLOs specified by the consumer.</t>
<t>IETF Network Slices are not, of themselves, TE constructs.
However, a network operator that offers IETF Network Slices is
likely to use many TE tools in order to manage their network and
provide the services.</t>
<t>IETF Network Slices are defined such that they are independent
of the underlying infrastructure connectivity and technologies
used. From a customer's perspective, an IETF Network Slice looks
like a VPN connectivity matrix with additional information about
the level of service that the customer requires between the
endpoints. From an operator's perspective, the IETF Network Slice
looks like a set of routing or tunneling instructions with the
network resource reservations necessary to provide the required
service levels as specified by the SLOs. The concept of an IETF
Network Slice is consistent with an enhanced VPN <xref
target="I-D.ietf-teas-enhanced-vpn" format="default"/>.</t>
</section>
</section>
<section anchor="TEtech" numbered="true" toc="default">
<name>IETF Techniques Used by TE Mechanisms</name>
<section anchor="CSPF" numbered="true" toc="default">
<name>Constraint-Based Routing</name>
<t>Constraint-based routing refers to a class of routing systems tha
t
compute routes through a network subject to the satisfaction of a set compute routes through a network subject to the satisfaction of a set
of constraints and requirements. In the most general case, of constraints and requirements. In the most general case,
constraint-based routing may also seek to optimize overall network constraint-based routing may also seek to optimize overall network
performance while minimizing costs.</t> performance while minimizing costs.</t>
<t>The constraints and requirements may be imposed by the network it
<t>The constraints and requirements may be imposed by the network itself self
or by administrative policies. Constraints may include bandwidth, or by administrative policies. Constraints may include bandwidth,
hop count, delay, and policy instruments such as resource class hop count, delay, and policy instruments such as resource class
attributes. Constraints may also include domain-specific attributes attributes. Constraints may also include domain-specific attributes
of certain network technologies and contexts which impose of certain network technologies and contexts that impose
restrictions on the solution space of the routing function. Path restrictions on the solution space of the routing function. Path-ori
oriented technologies such as MPLS have made constraint-based routing ented technologies such as MPLS have made constraint-based routing
feasible and attractive in public IP networks.</t> feasible and attractive in public IP networks.</t>
<t>The concept of constraint-based routing within the context of MPLS <t>The concept of constraint-based routing within the context of
TE requirements in IP networks was first described in MPLS-TE requirements in IP networks was first described in <xref
<xref target="RFC2702"/> and led to developments such as MPLS-TE target="RFC2702" format="default"/> and led to developments such
<xref target="RFC3209"/> as described in <xref target="MPLS"/>.</t> as MPLS-TE <xref target="RFC3209" format="default"/> as described
in <xref target="MPLS" format="default"/>.</t>
<t>Unlike QoS-based routing (for example, see <xref target="RFC2386"/>, <t>Unlike QoS-based routing (for example, see <xref
<xref target="MA"/>, and <xref target="I-D.ietf-idr-performance-routi target="RFC2386" format="default"/>, <xref target="MA"
ng"/>) format="default"/>, and <xref
which generally addresses the issue of routing individual traffic flo target="I-D.ietf-idr-performance-routing" format="default"/>) that
ws to generally addresses the issue of routing individual traffic flows
satisfy prescribed flow-based QoS requirements subject to network to satisfy prescribed flow-based QoS requirements subject to
resource availability, constraint-based routing is applicable to network resource availability, constraint-based routing is
traffic aggregates as well as flows and may be subject to a wide applicable to traffic aggregates as well as flows and may be
variety of constraints which may include policy restrictions.</t> subject to a wide variety of constraints that may include policy
restrictions.</t>
<section anchor="FLEX" title="IGP Flexible Algorithms (Flex-Algos)"> <section anchor="FLEX" numbered="true" toc="default">
<name>IGP Flexible Algorithms</name>
<t>The traditional approach to routing in an IGP network relies on th <t>The normal approach to routing in an IGP network relies
e IGPs deriving on the IGPs deriving "shortest paths" over the network based
"shortest paths" over the network based solely on the IGP metric a solely on the IGP metric assigned to the links. Such an
ssigned to approach is often limited: traffic may tend to converge
the links. Such an approach is often limited: traffic may tend to toward the destination, possibly causing congestion, and it is
converge toward not possible to steer traffic onto paths depending on the
the destination, possibly causing congestion; and it is not possib end-to-end qualities demanded by the applications.</t>
le to steer traffic <t>To overcome this limitation, various sorts of TE have been
onto paths depending on the end-to-end qualities demanded by the a widely deployed (as described in this document), where the TE
pplications.</t> component is responsible for computing the path based on
additional metrics and/or constraints. Such paths (or tunnels)
<t>To overcome this limitation, various sorts of TE have been widely need to be installed in the routers' forwarding tables in
deployed (as described in this document), where the TE component i addition to, or as a replacement for, the original paths computed
s responsible for by IGPs. The main drawbacks of these TE approaches are the
computing the path based on additional metrics and/or constraints. additional complexity of protocols and management and the state
Such paths (or that may need to be maintained within the network.</t>
tunnels) need to be installed in the routers&apos; forwarding tabl
es in addition to,
or as a replacement for the original paths computed by IGPs. The
main drawback of
these TE approaches is the additional complexity of protocols and
management, and
the state that may need to be maintained within the network.</t>
<t>IGP flexible algorithms (flex-algos) <xref target="RFC9350" /> all
ow IGPs
to construct constraint-based paths over the network by computing
constraint-
based next hops. The intent of flex-algos is to reduce TE complex
ity by letting an
IGP perform some basic TE computation capabilities. Flex-algo inc
ludes a set of
extensions to the IGPs that enable a router to send TLVs that:
<list style="symbols">
<t>describe a set of constraints on the topology</t>
<t>identify calculation-type</t>
<t>describe a metric-type that is to be used to compute the bes
t
paths through the constrained topology.</t>
</list>
A given combination of calculation-type, metric-type, and constrai
nts is known as a
"Flexible Algorithm Definition" (or FAD). A router that sends suc
h a set of TLVs also
assigns a specific identifier (the Flexible Algorithm) to the spec
ified combination of
calculation-type, metric-type, and constraints.</t>
<t>There are two use cases for flex-algo: in IP networks <xref target
="I-D.ietf-lsr-ip-flexalgo" />
and in Segment Routing networks <xref target="RFC9350" />. In the
first case,
flex-algo computes paths to an IPv4 or IPv6 address, in the second
case, flex-algo computes paths
to a prefix SID (see <xref target="SR" />).</t>
<t>Examples of where flex-algo can be useful include:
<list style="symbols">
<t>Expansion of the function of IP Performance metrics <xref ta
rget="RFC5664" />
where specific constraint-based routing (flex-algo) can be i
nstantiated within the
network based on the results of performance measurement.</t>
<t>The formation of an 'underlay' network using flex-algo, and
the realization of
an 'overlay' network using TE techniques. This approach can
leverage the nested
combination of flex-algo and TE extensions for IGP (see <xre
f target="IGPTE" />).</t>
<t>Flex-algo in SR-MPLS (<xref target="SR" />) can be used as a
base to easily
build a TE-like topology without TE components on routers or
the use of a PCE
(see <xref target="PCE" />).</t>
<t>The support for network slices <xref target="I-D.ietf-teas-i
etf-network-slices" />
where the SLOs of a particular IETF network slice can be gua
ranteed by a flex-algo,
or where a Filtered Topology <xref target="I-D.ietf-teas-iet
f-network-slices" />
can be created as a TE-like topology using a flex-algo.</t>
</list></t>
</section>
</section>
<section anchor="RSVP" title="RSVP">
<t>RSVP is a soft-state signaling protocol <xref target="RFC2205"/>. It
supports
receiver-initiated establishment of resource reservations for both
multicast and unicast flows. RSVP was originally developed as a
signaling protocol within the Integrated Services framework (see
<xref target="INTSERV" />) for applications to communicate QoS requir
ements
to the network and for the network to reserve relevant resources to s
atisfy
the QoS requirements <xref target="RFC2205"/>.</t>
<t>In RSVP, the traffic sender or source node sends a PATH message to th
e
traffic receiver with the same source and destination addresses as th
e
traffic which the sender will generate. The PATH message contains:
(1) a sender traffic specification describing the characteristics of
the
traffic, (2) a sender template specifying the format of the traffic,
and
(3) an optional advertisement specification which is used to support
the
concept of One Pass With Advertising (OPWA) <xref target="RFC2205"/>.
Every intermediate router along the path forwards the PATH message to
the
next hop determined by the routing protocol. Upon receiving a PATH m
essage,
the receiver responds with a RESV message which includes a flow descr
iptor
used to request resource reservations. The RESV message travels to t
he
sender or source node in the opposite direction along the path that
the PATH message traversed. Every intermediate router along the path
can reject or accept the reservation request of the RESV message. If
the request is rejected, the rejecting router will send an error
message to the receiver and the signaling process will terminate. If
the request is accepted, link bandwidth and buffer space are
allocated for the flow and the related flow state information is
installed in the router.</t>
<t>One of the issues with the original RSVP specification was
scalability. This was because reservations were required for micro-
flows, so that the amount of state maintained by network elements
tended to increase linearly with the number of traffic flows. These
issues are described in <xref target="RFC2961"/> which also modifies
and extends RSVP to mitigate the scaling problems to make RSVP a
versatile signaling protocol for the Internet. For example, RSVP has
been extended to reserve resources for aggregation of flows <xref tar
get="RFC3175" />, to set
up MPLS explicit label switched paths (see <xref target="MPLS" />),
and to perform other signaling functions within the Internet.
<xref target="RFC2961"/> also describes a mechanism to reduce the
amount of Refresh messages required to maintain established RSVP
sessions.</t>
</section>
<section anchor="MPLS" title="Multiprotocol Label Switching (MPLS)">
<t>MPLS is a forwarding scheme which also includes extensions
to conventional IP control plane protocols. MPLS extends the
Internet routing model and enhances packet forwarding and path
control <xref target="RFC3031"/>.</t>
<t>At the ingress to an MPLS domain, Label Switching Routers (LSRs)
classify IP packets into Forwarding Equivalence Classes (FECs) based
on a variety of factors, including, e.g., a combination of the
information carried in the IP header of the packets and the local
routing information maintained by the LSRs. An MPLS label stack entr
y
is then prepended to each packet according to their forwarding equiva
lence
classes. The MPLS label stack entry is 32 bits long and contains a 2
0-bit
label field.</t>
<t>An LSR makes forwarding decisions by using the label prepended to
packets as the index into a local next hop label forwarding entry
(NHLFE). The packet is then processed as specified in the NHLFE.
The incoming label may be replaced by an outgoing label (label swap),
and the packet may be forwarded to the next LSR. Before a packet
leaves an MPLS domain, its MPLS label may be removed (label pop). A
Label Switched Path (LSP) is the path between an ingress LSR and an
egress LSR through which a labeled packet traverses. The path of an
explicit LSP is defined at the originating (ingress) node of the LSP.
MPLS can use a signaling protocol such as RSVP or the Label Distribut
ion
Protocol (LDP) to set up LSPs.</t>
<t>MPLS is a powerful technology for Internet TE
because it supports explicit LSPs which allow constraint-based
routing to be implemented efficiently in IP networks <xref target="AW
D2"/>. The
requirements for TE over MPLS are described in
<xref target="RFC2702"/>. Extensions to RSVP to support instantiatio
n of explicit
LSP are discussed in <xref target="RFC3209"/> and <xref target="RSVP-
TE"/>.</t>
</section>
<section anchor="RSVP-TE" title="RSVP-TE">
<t>RSVP-TE is a protocol extension of RSVP (<xref target="RSVP"/>) for t
raffic engineering.
The base specification is found in <xref target="RFC3209"/>. RSVP-TE
enables the
establishment of traffic-engineered MPLS LSPs (TE LSPs), using loose
or strict paths,
and taking into consideration network constraints such as available b
andwidth. The
extension supports signaling LSPs on explicit paths that could be adm
inistratively
specified, or computed by a suitable entity (such as a PCE, <xref tar
get="PCE"/>)
based on QoS and policy requirements, taking into consideration the p
revailing network
state as advertised by IGP extension for IS-IS in <xref target="RFC53
05" />,
for OSPFV2 in <xref target="RFC3630" />, and for OSPFv3 in
<xref target="RFC5329" />. RSVP-TE enables the reservation of resour
ces
(for example, bandwidth) along the path.</t>
<t>RSVP-TE includes the ability to preempt LSPs based on priorities, and
uses link affinities
to include or exclude links from the LSPs. The protocol is further e
xtended to
support Fast Reroute (FRR) <xref target="RFC4090"/>, Diffserv <xref t
arget="RFC4124"/>,
and bidirectional LSPs <xref target="RFC7551"/>. RSVP-TE extensions
for support for
GMPLS (see <xref target="GMPLS"/>) are specified in <xref target="RFC
3473"/>.</t>
<t>Requirements for point-to-multipoint (P2MP) MPLS TE LSPs are document
ed in
<xref target="RFC4461"/>, and signaling protocol extensions for setti
ng up P2MP MPLS TE
LSPs via RSVP-TE are defined in <xref target="RFC4875"/> where a P2MP
LSP comprise
multiple source-to-leaf (S2L) sub-LSPs. To determine the paths for P
2MP LSPs,
selection of the branch points (based on capabilities, network state,
and policies) is
key <xref target="RFC5671" /></t>
<t>RSVP-TE has evolved to provide real time dynamic metrics for path sel
ection for low latency
paths using extensions to IS-IS <xref target="RFC8570" /> and OSPF
<xref target="RFC7471" /> based on STAMP <xref target="RFC8972" />
and TWAMP <xref target="RFC5357" /> performance measurements.</t>
<t>RSVP-TE has historically been used when bandwidth was constrained, ho
wever, as bandwidth
has increased, RSVP-TE has developed into a bandwidth management tool
to provide bandwidth
efficiency and proactive resource management.</t>
</section>
<section anchor="GMPLS" title="Generalized MPLS (GMPLS)">
<t>GMPLS extends MPLS control protocols to encompass time-division (e.g.
,
Synchronous Optical Network / Synchronous Digital Hierarchy (SONET/SD
H),
Plesiochronous Digital Hierarchy (PDH), Optical Transport Network (OT
N)),
wavelength (lambdas), and spatial switching (e.g., incoming port or f
iber
to outgoing port or fiber) as well as continuing to support packet sw
itching.
GMPLS provides a common set of control protocols for all of these lay
ers
(including some technology-specific extensions) each of which has a d
istinct
data or forwarding plane. GMPLS covers both the signaling and the ro
uting
part of that control plane and is based on the TE extensions
to MPLS (see <xref target="RSVP-TE"/>).</t>
<t>In GMPLS <xref target="RFC3945" />, the original MPLS architecture is
extended to include LSRs whose
forwarding planes rely on circuit switching, and therefore cannot for
ward
data based on the information carried in either packet or cell header
s.
Specifically, such LSRs include devices where the switching is based
on
time slots, wavelengths, or physical ports. These additions impact
basic LSP properties: how labels are requested and communicated, the
unidirectional nature of MPLS LSPs, how errors are propagated, and
information provided for synchronizing the ingress and egress LSRs <x
ref target="RFC3473" />.</t>
</section>
<section anchor="IPPM" title="IP Performance Metrics">
<t>The IETF IP Performance Metrics (IPPM) working group has developed a
set
of standard metrics that can be used to monitor the quality, performa
nce,
and reliability of Internet services. These metrics can be applied b
y
network operators, end-users, and independent testing groups to provi
de
users and service providers with a common understanding of the perfor
mance
and reliability of the Internet component &apos;clouds&apos; they use
/provide
<xref target="RFC2330"/>. The criteria for performance metrics devel
oped by
the IPPM working group are described in <xref target="RFC2330"/>. Ex
amples
of performance metrics include one-way packet loss <xref target="RFC7
680"/>,
one-way delay <xref target="RFC7679"/>, and connectivity measures bet
ween
two nodes <xref target="RFC2678"/>. Other metrics include second-ord
er
measures of packet loss and delay.</t>
<t>Some of the performance metrics specified by the IPPM working group a
re useful
for specifying SLAs. SLAs are sets of service level objectives negot
iated
between users and service providers, wherein each objective is a comb
ination
of one or more performance metrics, possibly subject to certain const
raints.</t>
<t>The IPPM working group also designs measurement techniques and protoc
ols to obtain
thwse metrics.</t>
</section>
<section anchor="RTFM" title="Flow Measurement">
<t>The IETF Real Time Flow Measurement (RTFM) working group produced
an architecture that defines a method to specify traffic flows
as well as a number of components for flow measurement (meters, meter
readers, manager) <xref target="RFC2722"/>. A flow measurement syste
m enables
network traffic flows to be measured and analyzed at the flow level
for a variety of purposes. As noted in RFC 2722, a flow measurement
system can be very useful in the following contexts:
<list style="symbols">
<t>understanding the behavior of existing networks</t>
<t>planning for network development and expansion</t>
<t>quantification of network performance</t>
<t>verifying the quality of network service</t>
<t>attribution of network usage to users.</t>
</list></t>
<t>A flow measurement system consists of meters, meter readers, and
managers. A meter observes packets passing through a measurement
point, classifies them into groups, accumulates usage data (such as
the number of packets and bytes for each group), and stores the usage
data in a flow table. A group may represent any collection of user
applications, hosts, networks, etc. A meter reader gathers usage dat
a
from various meters so it can be made available for analysis. A mana
ger
is responsible for configuring and controlling meters and meter reade
rs.
The instructions received by a meter from a manager include flow
specifications, meter control parameters, and sampling techniques. T
he
instructions received by a meter reader from a manager include the ad
dress
of the meter whose data are to be collected, the frequency of data co
llection,
and the types of flows to be collected.</t>
<t>IP Flow Information Export (IPFIX) <xref target="RFC5470" /> defines
an
architecture that is very similar to the RTFM architecture and includ
es
Metering, Exporting, and Collecting Processes. <xref target="RFC5472
" />
describes the applicability of IPFIX and makes a comparison with RTFM
, pointing
out that, architecturally, while RTM talks about devices, IPFIX deals
with processed
to clarify that multiple of those processes may be co-located on the
same machine.
The IPFIX protocol <xref target="RFC7011" /> is widely implemented.</
t>
</section>
<section anchor="ECM" title="Endpoint Congestion Management">
<t><xref target="RFC3124" /> provides a set of congestion control mechan
isms for the
use of transport protocols. It also allows the development of mechan
isms for
unifying congestion control across a subset of an endpoint&apos;s act
ive unicast
connections (called a congestion group). A congestion manager contin
uously
monitors the state of the path for each congestion group under its co
ntrol. The
manager uses that information to instruct a scheduler on how to parti
tion bandwidth
among the connections of that congestion group.</t>
<t>The concepts described in <xref target="RFC3124" /> and the lessons t
hat can be learned
from that work found a home in HTTP/2 <xref target="RFC9113" /> and Q
UIC <xref target="RFC9000" />,
while <xref target="RFC9040" /> describes TCP control block interdepe
ndence which is a
core construct underpinning the congestion manager defined in <xref t
arget="RFC3124" />.</t>
</section>
<section anchor="IGPTE" title="TE Extensions to the IGPs">
<t><xref target="RFC5305" /> describes the extensions to the Intermediat
e System to
Intermediate System (IS-IS) protocol to support TE, similarly <xref t
arget="RFC3630" />
specifies TE extensions for OSPFv2, and <xref target="RFC5329" /> has
the same description
for OSPFv3.</t>
<t>IS-IS and OSPF share the common concept of TE extensions to distribut
e TE parameters such as link
type and ID, local and remote IP addresses, TE metric, maximum bandwi
dth, maximum reservable
bandwidth and unreserved bandwidth, and admin group. The information
distributed by the IGPs in this way can be used to build a view of th
e state and capabilities
of a TE network (see <xref target="STATE" />).</t>
<t>The difference between IS-IS and OSPF is in the details of how they e
ncode and transmit the
TE parameters:
<list style="symbols">
<t>IS-IS uses the Extended IS Reachability TLV (type 22), the Exten
ded IP Reachability TLV
(type 135), and the TE Router ID TLV (type 134). These TLVs use
specific Sub-TLVs
described in <xref target="RFC8570" /> to carry for the TE param
eters.</t>
<t>OSPFv2 uses Opaque LSA <xref target="RFC5250" /> type 10 and OSP
Fv3 uses the Intra-Area-TE-LSA.
In both OSPF cases, two top-level TLVs are used (Router Address
and Link TLVs), and these
use Sub-TLVs to carry the TE parameters (as defined in <xref tar
get="RFC7471" /> for OSPFv2
and <xref target="RFC5329" /> for OSPFv3.</t>
</list></t>
</section>
<section anchor="BGPLS" title="BGP Link-State">
<t>In a number of environments, a component external to a network is ca
lled
upon to perform computations based on the network topology and curre
nt
state of the connections within the network, including TE
information. This is information typically distributed by IGP routi
ng
protocols within the network (see <xref target="IGPTE" />).</t>
<t>The Border Gateway Protocol (BGP) (see also <xref target="INTER" />)
is one of the
essential routing protocols that glue the Internet together. BGP Li
nk
State (BGP-LS) <xref target="RFC7752" /> is a mechanism by which
link-state and TE information can be collected from
networks and shared with external components using the BGP routing
protocol. The mechanism is applicable to physical and virtual IGP
links, and is subject to policy control.</t>
<t>Information collected by BGP-LS can be used, for example, to constru
ct the TED
(<xref target="STATE" />) for use by the
Path Computation Element (PCE, see <xref target="PCE" />), or may be
used
by Application-Layer Traffic Optimization (ALTO) servers (see
<xref target="ALTO" />).</t>
</section> <t>IGP Flexible Algorithms <xref target="RFC9350"
format="default"/> allow IGPs to construct constraint-based
paths over the network by computing constraint-based next hops.
The intent of Flexible Algorithms is to reduce TE complexity by le
tting
an IGP perform some basic TE computation capabilities.
Flexible Algorithm includes a set of extensions to the IGPs that e
nable a
router to send TLVs that:</t>
<ul spacing="normal">
<li>describe a set of constraints on the topology</li>
<li>identify calculation-type</li>
<li>describe a metric-type that is to be used to compute the
best paths through the constrained topology</li>
</ul>
<t>A given combination of calculation-type, metric-type, and
constraints is known as a Flexible Algorithm Definition
(FAD). A router that sends such a set of TLVs also assigns a
specific identifier (the Flexible Algorithm) to the specified
combination of calculation-type, metric-type, and
constraints.</t>
<t>There are two use cases for Flexible Algorithm: in IP networks
<xref
target="RFC9502" format="default"/> and in SR
networks <xref target="RFC9350" format="default"/>. In the
first case, Flexible Algorithm computes paths to an IPv4 or IPv6 a
ddress;
in the second case, Flexible Algorithms computes paths to a Prefix
SID
(see <xref target="SR" format="default"/>).</t>
<t>Examples of where Flexible Algorithms can be useful include:</t
>
<ul spacing="normal">
<li>Expansion of the function of IP performance metrics <xref
target="RFC5664" format="default"/> where specific
constraint-based routing (Flexible Algorithm) can be instantiate
d
within the network based on the results of performance
measurement.</li>
<li>The formation of an "underlay" network using Flexible Algori
thms,
and the realization of an "overlay" network using TE
techniques. This approach can leverage the nested combination
of Flexible Algorithm and TE extensions for IGP (see <xref
target="IGPTE" format="default"/>).</li>
<li>Flexible Algorithms in SR-MPLS (<xref target="SR"
format="default"/>) can be used as a base to easily build a
TE-like topology without TE components on routers or the use
of a PCE (see <xref target="PCE" format="default"/>).</li>
<section anchor="PCE" title="Path Computation Element"> <li>The support for network slices <xref
target="I-D.ietf-teas-ietf-network-slices" format="default"/>
where the SLOs of a particular IETF Network Slice can be
guaranteed by a Flexible Algorithm or where a Filtered Topology
<xref
target="I-D.ietf-teas-ietf-network-slices" format="default"/>
can be created as a TE-like topology using a Flexible Algorithm.
</li>
</ul>
</section>
</section>
<section anchor="RSVP" numbered="true" toc="default">
<name>RSVP</name>
<t>RSVP is a soft-state signaling protocol <xref target="RFC2205"
format="default"/>. It supports receiver-initiated establishment
of resource reservations for both multicast and unicast flows.
RSVP was originally developed as a signaling protocol within the
Integrated Services framework (see <xref target="INTSERV"
format="default"/>) for applications to communicate QoS
requirements to the network and for the network to reserve
relevant resources to satisfy the QoS requirements <xref
target="RFC2205" format="default"/>.</t>
<t>In RSVP, the traffic sender or source node sends a Path message
to the traffic receiver with the same source and destination
addresses as the traffic that the sender will generate. The Path
message contains:</t>
<ul spacing="normal">
<li>A sender traffic specification describing the
characteristics of the traffic</li>
<li>A sender template specifying the format of the traffic</li>
<li>An optional advertisement specification that is used
to support the concept of One Pass With Advertising (OPWA) <xref
target="RFC2205" format="default"/></li>
</ul>
<t>Every intermediate router along the path forwards the Path
message to the next hop determined by the routing protocol. Upon
receiving a Path message, the receiver responds with a Resv
message that includes a flow descriptor used to request resource
reservations. The Resv message travels to the sender or source
node in the opposite direction along the path that the Path
message traversed. Every intermediate router along the path can
reject or accept the reservation request of the Resv message. If
the request is rejected, the rejecting router will send an error
message to the receiver, and the signaling process will terminate.
If the request is accepted, link bandwidth and buffer space are
allocated for the flow, and the related flow state information is
installed in the router.</t>
<t>One of the issues with the original RSVP specification <xref
target="RFC2205" format="default"/> was scalability. This was
because reservations were required for micro-flows, so that the
amount of state maintained by network elements tended to increase
linearly with the number of traffic flows. These issues are
described in <xref target="RFC2961" format="default"/>, which also
modifies and extends RSVP to mitigate the scaling problems to make
RSVP a versatile signaling protocol for the Internet. For
example, RSVP has been extended to reserve resources for
aggregation of flows <xref target="RFC3175" format="default"/>, to
set up MPLS explicit LSPs (see <xref target="MPLS"
format="default"/>), and to perform other signaling functions
within the Internet. <xref target="RFC2961" format="default"/>
also describes a mechanism to reduce the amount of Refresh
messages required to maintain established RSVP sessions.</t>
</section>
<section anchor="MPLS" numbered="true" toc="default">
<name>MPLS</name>
<t>MPLS is a forwarding scheme that also includes extensions to
conventional IP control plane protocols. MPLS extends the
Internet routing model and enhances packet forwarding and path
control <xref target="RFC3031" format="default"/>.</t>
<t>At the ingress to an MPLS domain,
LSRs classify IP packets into Forwarding Equivalence Classes
(FECs) based on a variety of factors, including, e.g., a
combination of the information carried in the IP header of the
packets and the local routing information maintained by the LSRs.
An MPLS label stack entry is then prepended to each packet
according to their FECs. The MPLS label
stack entry is 32 bits long and contains a 20-bit label field.</t>
<t>An LSR makes forwarding decisions by using the label prepended
to packets as the index into a local Next Hop Label Forwarding
Entry (NHLFE). The packet is then processed as specified in the
NHLFE. The incoming label may be replaced by an outgoing label
(label swap), and the packet may be forwarded to the next LSR.
Before a packet leaves an MPLS domain, its MPLS label may be
removed (label pop). An LSP is the path between an ingress LSR
and an egress LSR through which a labeled packet traverses. The
path of an explicit LSP is defined at the originating (ingress)
node of the LSP. MPLS can use a signaling protocol such as RSVP
or the Label Distribution Protocol (LDP) to set up LSPs.</t>
<t>MPLS is a powerful technology for Internet TE because it
supports explicit LSPs that allow constraint-based routing to be
implemented efficiently in IP networks <xref target="AWD2"
format="default"/>. The requirements for TE over MPLS are
described in <xref target="RFC2702" format="default"/>.
Extensions to RSVP to support instantiation of explicit LSP are
discussed in <xref target="RFC3209" format="default"/> and <xref
target="RSVP-TE" format="default"/>.</t>
</section>
<section anchor="RSVP-TE" numbered="true" toc="default">
<name>RSVP-TE</name>
<t>RSVP-TE is a protocol extension of RSVP (<xref target="RSVP"
format="default"/>) for traffic engineering. The base
specification is found in <xref target="RFC3209"
format="default"/>. RSVP-TE enables the establishment of
traffic-engineered MPLS LSPs (TE LSPs), using loose or strict
paths and taking into consideration network constraints such as
available bandwidth. The extension supports signaling LSPs on
explicit paths that could be administratively specified or
computed by a suitable entity (such as a PCE, <xref target="PCE"
format="default"/>) based on QoS and policy requirements, taking
into consideration the prevailing network state as advertised by the
IGP extension for IS-IS in <xref target="RFC5305"
format="default"/>, for OSPFv2 in <xref target="RFC3630"
format="default"/>, and for OSPFv3 in <xref target="RFC5329"
format="default"/>. RSVP-TE enables the reservation of resources
(for example, bandwidth) along the path.</t>
<t>RSVP-TE includes the ability to preempt LSPs based on
priorities and uses link affinities to include or exclude links
from the LSPs. The protocol is further extended to support Fast
Reroute (FRR) <xref target="RFC4090" format="default"/>, Diffserv
<xref target="RFC4124" format="default"/>, and bidirectional LSPs
<xref target="RFC7551" format="default"/>. RSVP-TE extensions for
support for GMPLS (see <xref target="GMPLS" format="default"/>)
are specified in <xref target="RFC3473" format="default"/>.</t>
<t>Requirements for point-to-multipoint (P2MP) MPLS-TE LSPs are
documented in <xref target="RFC4461" format="default"/>, and
signaling protocol extensions for setting up P2MP MPLS-TE LSPs via
RSVP-TE are defined in <xref target="RFC4875" format="default"/>,
where a P2MP LSP comprises multiple source-to-leaf (S2L) sub-LSPs.
To determine the paths for P2MP LSPs, selection of the branch
points (based on capabilities, network state, and policies) is
key <xref target="RFC5671" format="default"/></t>
<t>RSVP-TE has evolved to provide real-time dynamic metrics for
path selection for low-latency paths using extensions to IS-IS
<xref target="RFC8570" format="default"/> and OSPF <xref
target="RFC7471" format="default"/> based on performance
measurements for the Simple Two-Way Active
Measurement Protocol (STAMP) <xref target="RFC8972"
format="default"/> and the Two-Way Active Measurement Protocol (TWAM
P)
<xref target="RFC5357" format="default"/>.</t>
<t>RSVP-TE has historically been used when bandwidth was
constrained; however, as bandwidth has increased, RSVP-TE has
developed into a bandwidth management tool to provide bandwidth
efficiency and proactive resource management.</t>
</section>
<section anchor="GMPLS" numbered="true" toc="default">
<name>Generalized MPLS (GMPLS)</name>
<t>Constraint-based path computation is a fundamental building block fo <t>GMPLS extends MPLS control protocols to encompass time-division
r (e.g., Synchronous Optical Network / Synchronous Digital Hierarchy
TE in MPLS and GMPLS networks. Path computation in (SONET/SDH), Plesiochronous Digital Hierarchy (PDH), and Optical
Transport Network (OTN)), wavelength (lambdas), and spatial
switching (e.g., incoming port or fiber to outgoing port or fiber)
and continues to support packet switching. GMPLS
provides a common set of control protocols for all of these layers
(including some technology-specific extensions), each of which has
a distinct data or forwarding plane. GMPLS covers both the
signaling and the routing part of that control plane and is based
on the TE extensions to MPLS (see <xref target="RSVP-TE"
format="default"/>).</t>
<t>In GMPLS <xref target="RFC3945" format="default"/>, the
original MPLS architecture is extended to include LSRs whose
forwarding planes rely on circuit switching and therefore cannot
forward data based on the information carried in either packet or
cell headers. Specifically, such LSRs include devices where the
switching is based on time slots, wavelengths, or physical ports.
These additions impact basic LSP properties: how labels are
requested and communicated, the unidirectional nature of MPLS
LSPs, how errors are propagated, and information provided for
synchronizing the ingress and egress LSRs <xref target="RFC3473"
format="default"/>.</t>
</section>
<section anchor="IPPM" numbered="true" toc="default">
<name>IP Performance Metrics (IPPM)</name>
<t>The IETF IP Performance Metrics (IPPM) Working Group has
developed a set of standard metrics that can be used to monitor
the quality, performance, and reliability of Internet services.
These metrics can be applied by network operators, end users, and
independent testing groups to provide users and service providers
with a common understanding of the performance and reliability of
the Internet component clouds they use/provide <xref
target="RFC2330" format="default"/>. The criteria for performance
metrics developed by the IPPM Working Group are described in <xref
target="RFC2330" format="default"/>. Examples of performance
metrics include one-way packet loss <xref target="RFC7680"
format="default"/>, one-way delay <xref target="RFC7679"
format="default"/>, and connectivity measures between two nodes
<xref target="RFC2678" format="default"/>. Other metrics include
second-order measures of packet loss and delay.</t>
<t>Some of the performance metrics specified by the IPPM Working
Group are useful for specifying SLAs. SLAs are sets of SLOs
negotiated between users and service providers,
wherein each objective is a combination of one or more performance
metrics, possibly subject to certain constraints.</t>
<t>The IPPM Working Group also designs measurement techniques and
protocols to obtain these metrics.</t>
</section>
<section anchor="RTFM" numbered="true" toc="default">
<name>Flow Measurement</name>
<t>The IETF Real Time Flow Measurement (RTFM) Working Group
produced an architecture that defines a method to specify traffic
flows as well as a number of components for flow measurement
(meters, meter readers, and managers) <xref target="RFC2722"
format="default"/>. A flow measurement system enables network
traffic flows to be measured and analyzed at the flow level for a
variety of purposes. As noted in <xref target="RFC2722"
format="default"/>, a flow measurement system can be very useful
in the following contexts:
</t>
<ul spacing="normal">
<li>understanding the behavior of existing networks</li>
<li>planning for network development and expansion</li>
<li>quantification of network performance</li>
<li>verifying the quality of network service</li>
<li>attribution of network usage to users</li>
</ul>
<t>A flow measurement system consists of meters, meter readers,
and managers. A meter observes packets passing through a
measurement point, classifies them into groups, accumulates usage
data (such as the number of packets and bytes for each group), and
stores the usage data in a flow table. A group may represent any
collection of user applications, hosts, networks, etc. A meter
reader gathers usage data from various meters so it can be made
available for analysis. A manager is responsible for configuring
and controlling meters and meter readers. The instructions
received by a meter from a manager include flow specifications,
meter control parameters, and sampling techniques. The
instructions received by a meter reader from a manager include the
address of the meter whose data are to be collected, the frequency
of data collection, and the types of flows to be collected.</t>
<t>IP Flow Information Export (IPFIX) <xref target="RFC5470"
format="default"/> defines an architecture that is very similar to
the RTFM architecture and includes Metering, Exporting, and
Collecting Processes. <xref target="RFC5472" format="default"/>
describes the applicability of IPFIX and makes a comparison with
RTFM, pointing out that, architecturally, while RTM talks about
devices, IPFIX deals with processes to clarify that multiple of
those processes may be co-located on the same machine. The IPFIX
protocol <xref target="RFC7011" format="default"/> is widely
implemented.</t>
</section>
<section anchor="ECM" numbered="true" toc="default">
<name>Endpoint Congestion Management</name>
<t><xref target="RFC3124" format="default"/> provides a set of
congestion control mechanisms for the use of transport protocols.
It also allows the development of mechanisms for unifying
congestion control across a subset of an endpoint's active unicast
connections (called a "congestion group"). A congestion manager
continuously monitors the state of the path for each congestion
group under its control. The manager uses that information to
instruct a scheduler on how to partition bandwidth among the
connections of that congestion group.</t>
<t>The concepts described in <xref target="RFC3124"
format="default"/> and the lessons that can be learned from that
work found a home in HTTP/2 <xref target="RFC9113"
format="default"/> and QUIC <xref target="RFC9000"
format="default"/>, while <xref target="RFC9040"
format="default"/> describes TCP control block interdependence
that is a core construct underpinning the congestion manager
defined in <xref target="RFC3124" format="default"/>.</t>
</section>
<section anchor="IGPTE" numbered="true" toc="default">
<name>TE Extensions to the IGPs</name>
<t><xref target="RFC5305" format="default"/> describes the
extensions to the Intermediate System to Intermediate System
(IS-IS) protocol to support TE. Similarly, <xref target="RFC3630"
format="default"/> specifies TE extensions for OSPFv2, and <xref
target="RFC5329" format="default"/> has the same description for
OSPFv3.</t>
<t>IS-IS and OSPF share the common concept of TE extensions to
distribute TE parameters, such as link type and ID, local and
remote IP addresses, TE metric, maximum bandwidth, maximum
reservable bandwidth, unreserved bandwidth, and admin group.
The information distributed by the IGPs in this way can be used to
build a view of the state and capabilities of a TE network (see
<xref target="STATE" format="default"/>).</t>
<t>The difference between IS-IS and OSPF is in the details of how
they encode and transmit the TE parameters:</t>
<ul spacing="normal">
<li>IS-IS uses the Extended IS Reachability TLV (type 22), the
Extended IP Reachability TLV (type 135), and the Traffic
Engineering router ID TLV (type 134). These TLVs use specific
sub-TLVs described in <xref target="RFC8570" format="default"/>
to carry the TE parameters.</li>
<li>OSPFv2 uses Opaque LSA <xref target="RFC5250"
format="default"/> type 10, and OSPFv3 uses the
Intra-Area-TE-LSA. In both OSPF cases, two top-level TLVs are
used (Router Address and Link TLVs), and these use sub-TLVs to
carry the TE parameters (as defined in <xref target="RFC7471"
format="default"/> for OSPFv2 and <xref target="RFC5329"
format="default"/> for OSPFv3).</li>
</ul>
</section>
<section anchor="BGPLS" numbered="true" toc="default">
<name>BGP - Link State</name>
<t>In a number of environments, a component external to a network
is called upon to perform computations based on the network
topology and current state of the connections within the network,
including TE information. This is information typically
distributed by IGP routing protocols within the network (see <xref
target="IGPTE" format="default"/>).</t>
<t>BGP (see also <xref target="INTER" format="default"/>) is
one of the essential routing protocols that glues the Internet
together. BGP-LS <xref target="RFC9552" format="default"/> is a
mechanism by which link-state and TE information can be collected
from networks and shared with external components using the BGP
routing protocol. The mechanism is applicable to physical and
virtual IGP links and is subject to policy control.</t>
<t>Information collected by BGP-LS can be used, for example, to
construct the TED (<xref target="STATE" format="default"/>) for
use by the PCE (see <xref target="PCE"
format="default"/>) or may be used by ALTO servers (see <xref target
="ALTO"
format="default"/>).</t>
</section>
<section anchor="PCE" numbered="true" toc="default">
<name>Path Computation Element </name>
<t>Constraint-based path computation is a fundamental building
block for TE in MPLS and GMPLS networks. Path computation in
large, multi-domain networks is complex and may require special large, multi-domain networks is complex and may require special
computational components and cooperation between the elements in computational components and cooperation between the elements in
different domains. The Path Computation Element (PCE) <xref target= different domains. The PCE <xref target="RFC4655"
"RFC4655"/> format="default"/> is an entity (component, application, or
is an entity (component, application, or network node) that is capab network node) that is capable of computing a network path or route
le of based on a network graph and applying computational
computing a network path or route based on a network graph and apply constraints.</t>
ing <t>Thus, a PCE can provide a central component in a TE system
computational constraints.</t> operating on the TED (see <xref target="STATE"
format="default"/>) with delegated responsibility for determining
<t>Thus, a PCE can provide a central component in a TE system paths in MPLS, GMPLS, or SR networks. The PCE uses
operating on the TE Database (TED, see <xref target="STATE" />) the Path Computation Element Communication Protocol (PCEP) <xref
with delegated responsibility for determining paths in MPLS, GMPLS, target="RFC5440" format="default"/> to communicate with Path
or Computation Clients (PCCs), such as MPLS LSRs, to answer their
Segment Routing networks. The PCE uses the Path Computation Element requests for computed paths or to instruct them to initiate new
Communication paths <xref target="RFC8281" format="default"/> and maintain state
Protocol (PCEP) <xref target="RFC5440" /> to communicate with Path C about paths already installed in the network <xref
omputation target="RFC8231" format="default"/>.</t>
Clients (PCCs), such as MPLS LSRs, to answer their requests for comp <t>PCEs form key components of a number of TE systems. More
uted paths information about the applicability of PCEs can be found in <xref
or to instruct them to initiate new paths <xref target="RFC8281" /> target="RFC8051" format="default"/>, while <xref target="RFC6805"
and maintain format="default"/> describes the application of PCEs to determining
state about paths already installed in the network <xref target="RFC paths across multiple domains. PCEs also have potential uses in
8231" />.</t> Abstraction and Control of TE Networks (ACTN) (see <xref
target="ACTN" format="default"/>), Centralized Network Control
<t>PCEs form key components of a number of TE systems. More informatio <xref target="RFC8283" format="default"/>, and
n SDN (see <xref target="SDN"
about the applicability of PCE can be found in <xref target="RFC8051 format="default"/>).</t>
"/>, while </section>
<xref target="RFC6805" /> describes the application of PCE to determ <section anchor="SR" numbered="true" toc="default">
ining paths <name>Segment Routing (SR)</name>
across multiple domains. PCE also has potential use in Abstraction <t>The SR architecture <xref target="RFC8402" format="default"/>
and Control of leverages the source routing and tunneling paradigms. The path a
TE Networks (ACTN) (see <xref target="ACTN" />), Centralized Network packet takes is defined at the ingress, and the packet is tunneled
Control
<xref target="RFC8283" />, and Software Defined Networking (SDN) (se
e
<xref target="SDN" />).</t>
</section>
<section anchor="SR" title="Segment Routing">
<t>The Segment Routing (SR) architecture <xref target="RFC8402" /> leve
rages the source routing and tunneling paradigms.
The path a packet takes is defined at the ingress and the packet is
tunneled
to the egress.</t> to the egress.</t>
<t>In a protocol realization, an ingress node steers a packet
using a set of instructions, called "segments", that are included in
an SR header prepended to the packet: a label stack in MPLS case,
and a series of 128-bit SIDs in the IPv6 case.</t>
<t>Segments are identified by SIDs. There
are four types of SIDs that are relevant for TE.</t>
<ul>
<li>Prefix SID:
A SID that is unique within the routing domain and is used
to identify a prefix.</li>
<li>Node SID:
A Prefix SID with the "N" bit set to identify a node.</li>
<li>Adjacency SID:
Identifies a unidirectional adjacency.</li>
<li><t>Binding SID:
A Binding SID has two purposes:</t>
<ol spacing="normal" type="1">
<li>To advertise the mappings of prefixes to SIDs/Labels</li>
<li>To advertise a path available for a Forwarding Equivalence
Class (FEC)</li>
</ol></li>
</ul>
<t>A segment can represent any instruction, topological or
service-based. SIDs can be looked up in a global context
(domain-wide) as well as in some other contexts (see, for example,
"context labels" in <xref target="RFC5331" sectionFormat="of"
section="3"/>).</t>
<t>The application of policy to SR can make SR into
a TE mechanism, as described in <xref target="SRPolicy"
format="default"/>.</t>
</section>
<section anchor="BIER-TE" numbered="true" toc="default">
<name>Tree Engineering for Bit Index Explicit Replication
</name>
<t>Bit Index Explicit Replication (BIER) <xref target="RFC8279"
format="default"/> specifies an encapsulation for multicast
forwarding that can be used on MPLS or Ethernet transports. A
mechanism known as Tree Engineering for Bit Index Explicit
Replication (BIER-TE) <xref target="RFC9262" format="default"/>
provides a component that could be used to build a
traffic-engineered multicast system. BIER-TE does not of itself
offer full traffic engineering, and the abbreviation "TE" does
not, in this case, refer to traffic engineering.</t>
<t>In BIER-TE, path steering is supported via the definition of a
bitstring attached to each packet that determines how the packet
is forwarded and replicated within the network. Thus, this
bitstring steers the traffic within the network and forms an
element of a traffic-engineering system. A central controller
that is aware of the capabilities and state of the network as well
as the demands of the various traffic flows is able to select
multicast paths that take account of the available resources and
demands. Therefore, this controller is responsible for the
policy elements of traffic engineering.</t>
<t>Resource management has implications for the forwarding plane
beyond the steering of packets defined for BIER-TE. These include
the allocation of buffers to meet the requirements of admitted
traffic and may include policing and/or rate-shaping mechanisms
achieved via various forms of queuing. This level of resource
control, while optional, is important in networks that wish to
support congestion management policies to control or regulate the
offered traffic to deliver different levels of service and
alleviate congestion problems. It is also important in networks that
wish to control latencies experienced by specific traffic
flows.</t>
</section>
<section anchor="STATE" numbered="true" toc="default">
<name>Network TE State Definition and Presentation</name>
<t>The network states that are relevant to TE need to be stored in
the system and presented to the user. The TED is a
collection of all TE information about all TE nodes and TE links
in the network. It is an essential component of a TE system, such
as MPLS-TE <xref target="RFC2702" format="default"/> or GMPLS
<xref target="RFC3945" format="default"/>. In order to formally
define the data in the TED and to present the data to the user,
the data modeling language YANG <xref target="RFC7950"
format="default"/> can be used as described in <xref
target="RFC8795" format="default"/>.</t>
</section>
<section anchor="SYSMAN" numbered="true" toc="default">
<name>System Management and Control Interfaces</name>
<t>In a protocol realization, an ingress node steers a packet using a s <t>The TE control system needs to have a management interface that
et of instructions, is human-friendly and a control interface that is programmable for
called segments, that are included in an SR header prepended to the automation. The Network Configuration Protocol (NETCONF) <xref
packet: a label stack in MPLS target="RFC6241" format="default"/> and the RESTCONF protocol
case, and a series of 128-bit segment identifiers in the IPv6 case.< <xref target="RFC8040" format="default"/> provide programmable
/t> interfaces that are also human-friendly. These protocols use XML-
or JSON-encoded messages. When message compactness or protocol
<t>Segments are identified by Segment Identifiers (SIDs). There are fo bandwidth consumption needs to be optimized for the control
ur types interface, other protocols, such as Group Communication for the
of SID that are relevant for TE. Constrained Application Protocol (CoAP) <xref target="RFC7390"
format="default"/> or gRPC <xref target="GRPC" format="default"/>,
<list style="symbols"> are available, especially when the protocol messages are encoded
<t>Prefix SID: A SID that is unique within the routing domain and in a binary format. Along with any of these protocols, the data
is used to identify a prefix.</t> modeling language YANG <xref target="RFC7950" format="default"/>
<t>Node SID: A Prefix SID with the 'N' bit set to identify a node can be used to formally and precisely define the interface
.</t> data.</t>
<t>Adjacency SID: Identifies a unidirectional adjacency.</t> <t>PCEP <xref target="RFC5440" format="default"/> is another
<t>Binding SID: A Binding SID has two purposes: protocol that has evolved to be an option for the TE system
<list style="numbers"> control interface. PCEP messages are TLV based; they are not
<t>To advertise the mappings of prefixes to SIDs/Labels.</ defined by a data-modeling language such as YANG.</t>
t> </section>
<t>To advertise a path available for a Forwarding Equivale </section>
nce Class.</t>
</list></t>
</list></t>
<t>A segment can represent any instruction, topological or service-base
d.
SIDs can be looked up in a global context (domain wide) as well as i
n some other
context (see, for example, "context labels" in Section 3 of <xref ta
rget="RFC5331" />).</t>
<t>The application of "policy" to Segment Routing can make SR into a TE
mechanism as described
in <xref target="SRPolicy" />.</t>
</section>
<section anchor="BIER-TE" title="Bit Index Explicit Replication Tree Engin
eering">
<t>Bit Index Explicit Replication (BIER) <xref target="RFC8279"/> speci
fies an encapsulation for
multicast forwarding that can be used on MPLS or Ethernet transports
. A mechanism known as
Tree Engineering for Bit Index Explicit Replication (BIER-TE) <xref
target="RFC9262"/>
provides a component that could be used to build a traffic-engineere
d multicast system.
BIER-TE does not of itself offer full traffic engineering, and the a
bbreviation "TE" does not, in
this case, refer to traffic engineering.</t>
<t>In BIER-TE, path steering is supported via the definition of a bitst
ring attached to each packet
that determines how the packet is forwarded and replicated within th
e network. Thus, this
bitstring steers the traffic within the network and forms an element
of a traffic engineering
system. A central controller that is aware of the capabilities and
state of the network as
well as the demands of the various traffic flows, is able to select
multicast paths that
take account of the available resources and demands. This controlle
r, therefore, is responsible
for the policy elements of traffic engineering.</t>
<t>Resource management has implications for the forwarding plane beyond
the steering of packets
defined for BIER-TE. These include the allocation of buffers to mee
t the requirements of admitted
traffic, and may include policing and/or rate-shaping mechanisms ach
ieved via various forms of
queuing. This level of resource control, while optional, is importa
nt in networks that wish to
support congestion management policies to control or regulate the of
fered traffic to deliver different
levels of service and alleviate congestion problems, or those networ
ks that wish to control latencies
experienced by specific traffic flows.</t>
</section>
<section anchor="STATE" title="Network TE State Definition and Presentatio
n">
<t>The network states that are relevant to TE need to be
stored in the system and presented to the user. The Traffic
Engineering Database (TED) is a collection of all TE information
about all TE nodes and TE links in the network. It is
an essential component of a TE system, such as MPLS-TE <xref target="
RFC2702" />
or GMPLS <xref target ="RFC3945" />. In order to formally define the
data
in the TED and to present the data to the user, the data
modeling language YANG <xref target="RFC7950" /> can be used as descr
ibed in
<xref target="RFC8795" />.</t>
</section> </section>
<section anchor="CDN" numbered="true" toc="default">
<section anchor="SYSMAN" title="System Management and Control Interfaces"> <name>Content Distribution</name>
<t>The Internet is dominated by client-server interactions,
<t>The TE control system needs to have a management principally web traffic and multimedia streams, although in the
interface that is human-friendly and a control interface that is future, more sophisticated media servers may become dominant. The
programmable for automation. The Network Configuration Protocol (NET location and performance of major information servers have a
CONF) significant impact on the traffic patterns within the Internet as well
<xref target="RFC6241" /> or the RESTCONF Protocol <xref target="RFC8 as on the perception of service quality by end users.</t>
040" /> <t>A number of dynamic load-balancing techniques have been devised to
provide programmable interfaces that are also human-friendly. These improve the performance of replicated information servers. These
protocols use XML or JSON encoded messages. When message compactness techniques can cause spatial traffic characteristics to become more
or dynamic in the Internet because information servers can be dynamically
protocol bandwidth consumption needs to be optimized for the control picked based upon the location of the clients, the location of the
interface, other protocols, such as Group Communication for the Const servers, the relative utilization of the servers, the relative
rained performance of different networks, and the relative performance of
Application Protocol (CoAP) <xref target="RFC7390" /> or gRPC <xref t different parts of a network. This process of assignment of
arget="GRPC" />, are available, distributed servers to clients is called "traffic directing". It is an
especially when the protocol messages are encoded in a binary format. application-layer function.</t>
Along <t>Traffic-directing schemes that allocate servers in multiple
with any of these protocols, the data modeling language YANG geographically dispersed locations to clients may require empirical
<xref target="RFC7950" /> can be used to formally and precisely defin network performance statistics to make more effective decisions. In
e the the future, network measurement systems may need to provide this type
interface data.</t> of information.</t>
<t>When congestion exists in the network, traffic-directing and
<t>The Path Computation Element Communication Protocol (PCEP) traffic-engineering systems should act in a coordinated manner. This
<xref target="RFC5440" /> is another protocol that has evolved to be topic is for further study.</t>
an option <t>The issues related to location and replication of information
for the TE system control interface. The messages of PCEP are TLV-ba servers, particularly web servers, are important for Internet traffic
sed, engineering because these servers contribute a substantial proportion
not defined by a data modeling language such as YANG.</t> of Internet traffic.</t>
</section> </section>
</section> </section>
<section anchor="RECO" numbered="true" toc="default">
</section> <name>Recommendations for Internet Traffic Engineering</name>
<t>This section describes high-level recommendations for traffic
<section anchor="CDN" title="Content Distribution"> engineering in the Internet in general terms.</t>
<t>The recommendations describe the capabilities needed to solve a TE
<t>The Internet is dominated by client-server interactions, principally problem or to achieve a TE objective. Broadly speaking, these
Web traffic and multimedia streams, although in the future, more sophisti recommendations can be categorized as either functional or
cated media servers may non-functional recommendations: </t>
become dominant. The location and performance of major information <ul spacing="normal">
servers has a significant impact on the traffic patterns within the <li>Functional recommendations describe the functions that a traffic-eng
Internet as well as on the perception of service quality by end ineering system
users.</t> should perform. These functions are needed to realize TE objectives
by addressing traffic-engineering problems.</li>
<t>A number of dynamic load-balancing techniques have been devised to <li>Non-functional recommendations relate to the quality attributes or
improve the performance of replicated information servers. These state characteristics of a TE system. These recommendations may
techniques can cause spatial traffic characteristics to become more contain conflicting assertions and may sometimes be difficult to
dynamic in the Internet because information servers can be quantify precisely.</li>
dynamically picked based upon the location of the clients, the </ul>
location of the servers, the relative utilization of the servers, the <t>The subsections that follow first summarize the non-functional
relative performance of different networks, and the relative requirements and then detail the functional requirements.</t>
performance of different parts of a network. This process of <section anchor="HIGHOBJ" numbered="true" toc="default">
assignment of distributed servers to clients is called traffic <name>Generic Non-functional Recommendations</name>
directing. It is an application layer function.</t> <t>The generic non-functional recommendations for Internet traffic
engineering are listed in the paragraphs that follow. In a given
<t>Traffic directing schemes that allocate servers in multiple context, some of these recommendations may be critical while others
geographically dispersed locations to clients may require empirical may be optional. Therefore, prioritization may be required during the
network performance statistics to make more effective decisions. In development phase of a TE system to tailor it to a specific
the future, network measurement systems may need to provide this type operational context.</t>
of information.</t> <dl newline="false" spacing="normal">
<dt>Automation:</dt>
<t>When congestion exists in the network, traffic directing and traffic <dd>Whenever feasible, a TE system should automate as many TE
engineering systems should act in a coordinated manner. This topic functions as possible to minimize the amount of human effort needed
is for further study.</t> to analyze and control operational networks. Automation is
particularly important in large-scale public networks because of the
<t>The issues related to location and replication of information high cost of the human aspects of network operations and the high
servers, particularly web servers, are important for Internet traffic risk of network problems caused by human errors. Automation may
engineering because these servers contribute a substantial proportion additionally benefit from feedback from the network that indicates
of Internet traffic.</t> the state of network resources and the current load in the network.
Further, placing intelligence into components of the TE system could
</section> enable automation to be more dynamic and responsive to changes in
the network.</dd>
</section> <dt>Flexibility:</dt>
<dd>A TE system should allow for changes in optimization policy. In
<section anchor="RECO" title="Recommendations for Internet Traffic Engineering"> particular, a TE system should provide sufficient configuration
options so that a network administrator can tailor the system to a
<t>This section describes high-level recommendations for traffic particular environment. It may also be desirable to have both
engineering in the Internet in general terms.</t> online and offline TE subsystems that can be independently enabled
and disabled. TE systems that are used in multi-class networks
<t>The recommendations describe the capabilities needed to solve a should also have options to support class-based performance
TE problem or to achieve a TE evaluation and optimization.</dd>
objective. Broadly speaking, these recommendations can be <dt>Interoperability:</dt>
categorized as either functional or non-functional recommendations. <dd>Whenever feasible, TE systems and their components should be
developed with open standards-based interfaces to allow
<list style="symbols"> interoperation with other systems and components.</dd>
<t>Functional recommendations describe the functions that a traffic <dt>Scalability:</dt>
engineering system should perform. These functions are needed to <dd>Public networks continue to grow rapidly with respect to
realize TE objectives by addressing traffic network size and traffic volume. Therefore, to remain applicable as
engineering problems.</t> the network evolves, a TE system should be scalable. In particular,
a TE system should remain functional as the network expands with
<t>Non-functional recommendations relate to the quality attributes regard to the number of routers and links and with respect to the
or state characteristics of a TE system. These number of flows and the traffic volume. A TE system should have a
recommendations may contain conflicting assertions and may sometimes scalable architecture, should not adversely impair other functions
be difficult to quantify precisely.</t> and processes in a network element, and should not consume too many
</list></t> network resources when collecting and distributing state
information or when exerting control.</dd>
<t>The subsections that follow first summarize the non-functional <dt>Security:</dt>
requirements, and then detail the functional requirements.</t> <dd>Security is a critical consideration in TE systems. Such
systems typically exert control over functional aspects of the
<section anchor="HIGHOBJ" title="Generic Non-functional Recommendations"> network to achieve the desired performance objectives. Therefore,
adequate measures must be taken to safeguard the integrity of the TE
<t>The generic non-functional recommendations for Internet traffic system. Adequate measures must also be taken to protect the network
engineering are listed in the paragraphs that follow. In a given from vulnerabilities that originate from security breaches and other
context, some of these recommendations may be critical while others impairments within the TE system.</dd>
may be optional. Therefore, prioritization may be required during <dt>Simplicity:</dt>
the development phase of a TE system to tailor it <dd>A TE system should be as simple as possible. Simplicity in
to a specific operational context.</t> user interface does not necessarily imply that the TE system will
use naive algorithms. When complex algorithms and internal
<t><list style="hanging"> structures are used, the user interface should hide such
<t hangText='Automation:'> complexities from the network administrator as much as
Whenever feasible, a TE system should automate as many TE functions a possible.</dd>
s <dt>Stability:</dt>
possible to minimize the amount of human effort needed to analyze and <dd>Stability refers to the resistance of the network to oscillate
control (flap) in a disruptive manner from one state to another, which may
operational networks. Automation is particularly important in large- result in traffic being routed first one way and then another
scale public without satisfactory resolution of the underlying TE issues and
networks because of the high cost of the human aspects of network ope with continued changes that do not settle down. Stability is a very
rations and important consideration in TE systems that respond to changes in the
the high risk of network problems caused by human errors. Automation state of the network. State-dependent TE methodologies typically
may additionally include a trade-off between responsiveness and stability. It is
benefit from feedback from the network that indicates the state of ne strongly recommended that when a trade-off between responsiveness
twork and stability is needed, it should be made in favor of stability
resources and the current load in the network. Further, placing inte (especially in public IP backbone networks).</dd>
lligence into <dt>Usability:</dt>
components of the TE system could enable automation to be more dynami <dd>Usability is a human aspect of TE systems. It refers to the
c and responsive ease with which a TE system can be deployed and operated. In
to changes in the network.</t> general, it is desirable to have a TE system that can be readily
deployed in an existing network. It is also desirable to have a TE
<t hangText='Flexibility:'> system that is easy to operate and maintain.</dd>
A TE system should allow for changes in optimization policy. In part <dt>Visibility:</dt>
icular, a <dd>Mechanisms should exist as part of the TE system to collect
TE system should provide sufficient configuration options so that a n statistics from the network and to analyze these statistics to
etwork determine how well the network is functioning. Derived statistics
administrator can tailor the system to a particular environment. It (such as traffic matrices, link utilization, latency, packet loss,
may also be and other performance measures of interest) that are determined from
desirable to have both online and offline TE subsystems which can be network measurements can be used as indicators of prevailing network
independently conditions. The capabilities of the various components of the
enabled and disabled. TE systems that are used in multi-class networ routing system are other examples of status information that should
ks should also be observable.</dd>
have options to support class-based performance evaluation and optimi </dl>
zation.</t> </section>
<section anchor="ROUTEREC" numbered="true" toc="default">
<t hangText='Interoperability:'> <name>Routing Recommendations</name>
Whenever feasible, TE systems and their components should be develope <t>Routing control is a significant aspect of Internet traffic
d with open
standards-based interfaces to allow interoperation with other systems
and components.</t>
<t hangText='Scalability:'>
Public networks continue to grow rapidly with respect to network size
and
traffic volume. Therefore, to remain applicable as the network evolv
es, a
TE system should be scalable. In particular, a TE system should rema
in
functional as the network expands with regard to the number of router
s and
links, and with respect to the number of flows and the traffic volume
. A TE system should have a
scalable architecture, should not adversely impair other functions an
d
processes in a network element, and should not consume too many netwo
rk
resources when collecting and distributing state information, or when
exerting control.</t>
<t hangText='Security:'>
Security is a critical consideration in TE systems. Such systems typ
ically exert
control over functional aspects of the network to achieve the desired
performance
objectives. Therefore, adequate measures must be taken to safeguard
the integrity
of the TE system. Adequate measures must also be taken to protect th
e network from
vulnerabilities that originate from security breaches and other impai
rments within
the TE system.</t>
<t hangText='Simplicity:'>
A TE system should be as simple as possible. Simplicity in user inte
rface does
not necessarily imply that the TE system will use naive algorithms.
When complex
algorithms and internal structures are used, the user interface shoul
d hide such
complexities from the network administrator as much as possible.</t>
<t hangText='Stability:'>
Stability refers to the resistance of the network to oscillate (flap
) in a disruptive
manner from one state to another, which may result in traffic being
routed first one
way and then another without satisfactory resolution of the underlyi
ng TE issues, and
with continued changes that do not settle down. Stability is a very
important
consideration in TE systems that respond to changes in the state of
the network.
State-dependent TE methodologies typically include a trade-off betwe
en responsiveness
and stability. It is strongly recommended that when a trade-off bet
ween responsiveness
and stability is needed, it should be made in favor of stability (es
pecially in public
IP backbone networks).</t>
<t hangText='Usability:'>
Usability is a human aspect of TE systems. It
refers to the ease with which a TE system can be
deployed and operated. In general, it is desirable to have a TE
system that can be readily deployed in an existing network. It is
also desirable to have a TE system that is easy to operate and mainta
in.</t>
<t hangText='Visibility:'>
Mechanisms should exist as part of the TE system to collect statistic
s from the
network and to analyze these statistics to determine how well the net
work is
functioning. Derived statistics such as traffic matrices, link utili
zation,
latency, packet loss, and other performance measures of interest whic
h are
determined from network measurements can be used as indicators of pre
vailing
network conditions. The capabilities of the various components of th
e routing
system are other examples of status information which should be obser
vable.</t>
</list></t>
</section>
<section anchor="ROUTEREC" title="Routing Recommendations">
<t>Routing control is a significant aspect of Internet traffic
engineering. Routing impacts many of the key performance measures engineering. Routing impacts many of the key performance measures
associated with networks, such as throughput, delay, and utilization. associated with networks, such as throughput, delay, and utilization.
Generally, it is very difficult to provide good service quality in a Generally, it is very difficult to provide good service quality in a
wide area network without effective routing control. A desirable TE wide area network without effective routing control. A desirable TE
routing system is one that takes traffic characteristics and network routing system is one that takes traffic characteristics and network
constraints into account during route selection while maintaining constraints into account during route selection while maintaining
stability.</t> stability.</t>
<t>Shortest Path First (SPF) IGPs are based on shortest path
algorithms and have limited control capabilities for TE <xref
target="RFC2702" format="default"/> <xref target="AWD2"
format="default"/>. These limitations include:</t>
<ol spacing="normal" type="1">
<li><t>Pure SPF protocols do not take network constraints and
traffic characteristics into account during route selection. For
example, IGPs always select the shortest paths based on link metrics
assigned by administrators, so load sharing cannot be performed
across paths of different costs. Note that link metrics are
assigned following a range of operator-selected policies that might
reflect preference for the use of some links over others; therefore,
"shortest" might not be purely a measure of distance. Using
shortest paths to forward traffic may cause the following
problems:</t>
<ul spacing="normal">
<li>If traffic from a source to a destination exceeds the
capacity of a link along the shortest path, the link (and hence
the shortest path) becomes congested while a longer path between
these two nodes may be under-utilized.</li>
<li>The shortest paths from different sources can overlap at
some links. If the total traffic from the sources exceeds the
capacity of any of these links, congestion will occur.</li>
<li>Problems can also occur because traffic demand changes over
time, but network topology and routing configuration cannot be
changed as rapidly. This causes the network topology and
routing configuration to become sub-optimal over time, which may
result in persistent congestion problems.</li>
</ul>
</li>
<li>The Equal-Cost Multipath (ECMP) capability of SPF IGPs supports
sharing of traffic among equal-cost paths. However, ECMP attempts
to divide the traffic as equally as possible among the equal-cost
shortest paths. Generally, ECMP does not support configurable
load-sharing ratios among equal-cost paths. The result is that one
of the paths may carry significantly more traffic than other paths
because it may also carry traffic from other sources. This
situation can result in congestion along the path that carries more
traffic. Weighted ECMP (WECMP) (see, for example, <xref
target="I-D.ietf-bess-evpn-unequal-lb" format="default"/>) provides
some mitigation.</li>
<li>Modifying IGP metrics to control traffic routing tends to have
network-wide effects. Consequently, undesirable and unanticipated
traffic shifts can be triggered as a result. Work described in
<xref target="PRACTICE" format="default"/> may be capable of better
control <xref target="FT00" format="default"/> <xref target="FT01"
format="default"/>.</li>
</ol>
<t>Because of these limitations, capabilities are needed to enhance
the routing function in IP networks. Some of these capabilities are
summarized below:</t>
<ul spacing="normal">
<li>Constraint-based routing computes routes to fulfill requirements
subject to constraints. This can be useful in public IP backbones
with complex topologies. Constraints may include bandwidth, hop
count, delay, and administrative policy instruments, such as resource
class attributes <xref target="RFC2702" format="default"/> <xref
target="RFC2386" format="default"/>. This makes it possible to
select routes that satisfy a given set of requirements. Routes
computed by constraint-based routing are not necessarily the
shortest paths. Constraint-based routing works best with
path-oriented technologies that support explicit routing, such as
MPLS.</li>
<li>Constraint-based routing can also be used as a way to distribute
traffic onto the infrastructure, including for best-effort traffic.
For example, congestion problems caused by uneven traffic
distribution may be avoided or reduced by knowing the reservable
bandwidth attributes of the network links and by specifying the
bandwidth requirements for path selection.</li>
<li>A number of enhancements to the link-state IGPs allow them to
distribute additional state information required for
constraint-based routing. The extensions to OSPF are described in
<xref target="RFC3630" format="default"/>, and the extensions to
IS-IS are described in <xref target="RFC5305" format="default"/>.
Some of the additional topology state information includes link
attributes, such as reservable bandwidth and link resource class
attribute (an administratively specified property of the link). The
resource class attribute concept is defined in <xref
target="RFC2702" format="default"/>. The additional topology state
information is carried in new TLVs and sub-TLVs in IS-IS <xref
target="RFC5305" format="default"/> or in the Opaque LSA in OSPF
<xref target="RFC3630" format="default"/>.</li>
<t>Shortest path first (SPF) IGPs are based on shortest path algorithms <li>An enhanced link-state IGP may flood information more frequently
and have limited control capabilities for TE <xref target="RFC2702"/>, than a normal IGP. This is because even without changes in
<xref target="AWD2"/>. These limitations include: topology, changes in reservable bandwidth or link affinity can
trigger the enhanced IGP to initiate flooding. A trade-off between
<list style="numbers"> the timeliness of the information flooded and the flooding frequency
<t>Pure SPF protocols do not take network constraints and traffic is typically implemented using a threshold based on the percentage
characteristics into account during route selection. For example, change of the advertised resources to avoid excessive consumption of
IGPs always select the shortest paths based on link metrics assigned link bandwidth and computational resources and to avoid instability
by administrators, so load sharing cannot be performed across paths in the TED.</li>
of different costs. Note that link metrics are assigned following a <li>In a TE system, it is also desirable for the routing subsystem
range of operator-selected policies that might reflect preference fo to make the load-splitting ratio among multiple paths (with equal
r cost or different cost) configurable. This capability gives network
the use of some links over others, and "shortest" might not, therefo
re,
be purely a measure of distance. Using shortest paths to forward tr
affic
may cause the following problems:
<list style="symbols">
<t>If traffic from a source to a destination exceeds the capacity
of a link along the shortest path, the link (and hence the shor
test
path) becomes congested while a longer path between these two
nodes may be under-utilized.</t>
<t>The shortest paths from different sources can overlap at some
links. If the total traffic from the sources exceeds the
capacity of any of these links, congestion will occur.</t>
<t>Problems can also occur because traffic demand changes over tim
e,
but network topology and routing configuration cannot be change
d
as rapidly. This causes the network topology and routing
configuration to become sub-optimal over time, which may result
in
persistent congestion problems.</t>
</list></t>
<t>The Equal-Cost Multi-Path (ECMP) capability of SPF IGPs supports
sharing of traffic among equal-cost paths.
However, ECMP attempts to divide the traffic as equally as
possible among the equal-cost shortest paths. Generally, ECMP
does not support configurable load sharing ratios among equal cost
paths. The result is that one of the paths may carry
significantly more traffic than other paths because it may also
carry traffic from other sources. This situation can result in
congestion along the path that carries more traffic. Weighted ECMP
(WECMP) (see, for example, <xref target="I-D.ietf-bess-evpn-unequal-
lb" />)
provides some mitigation.</t>
<t>Modifying IGP metrics to control traffic routing tends to have
network-wide effects. Consequently, undesirable and unanticipated
traffic shifts can be triggered as a result. Work
described in <xref target="PRACTICE"/> may be capable of better
control <xref target="FT00"/>, <xref target="FT01"/>.</t>
</list></t>
<t>Because of these limitations, capabilities are needed to enhance
the routing function in IP networks. Some of these capabilities are
summarized below.</t>
<t><list style="symbols">
<t>Constraint-based routing computes routes to fulfill requirements
subject to constraints. This can be useful in public IP backbones wit
h
complex topologies. Constraints may include bandwidth, hop count, del
ay,
and administrative policy instruments such as resource class attribute
s
<xref target="RFC2702"/>, <xref target="RFC2386"/>. This makes it pos
sible
to select routes that satisfy a given set of requirements. Routes com
puted
by constraint-based routing are not necessarily the shortest paths.
Constraint-based routing works best with path-oriented technologies th
at
support explicit routing, such as MPLS.</t>
</list></t>
<t><list style="none">
<t>Constraint-based routing can also be used as a way to distribute traff
ic
onto the infrastructure, including for best effort traffic. For examp
le,
congestion problems caused by uneven traffic distribution may be avoid
ed
or reduced by knowing the reservable bandwidth attributes of the netwo
rk
links and by specifying the bandwidth requirements for path selection.
</t>
</list></t>
<t><list style="symbols">
<t>A number of enhancements to the link state IGPs allow them
to distribute additional state information required for constraint-bas
ed
routing. The extensions to OSPF are described in <xref target="RFC363
0"/>,
and to IS-IS in <xref target="RFC5305"/>. Some of the additional topo
logy
state information includes link attributes such as reservable bandwidt
h and
link resource class attribute (an administratively specified property
of
the link). The resource class attribute concept is defined in
<xref target="RFC2702"/>. The additional topology state information i
s
carried in new TLVs and sub-TLVs in IS-IS, or in the Opaque LSA in OSP
F
<xref target="RFC5305"/>, <xref target="RFC3630"/>.</t>
</list></t>
<t><list style="none">
<t>An enhanced link-state IGP may flood information more frequently than
a normal IGP. This is because even without changes in topology,
changes in reservable bandwidth or link affinity can trigger the
enhanced IGP to initiate flooding. A trade-off between the timeliness
of
the information flooded and the flooding frequency is typically implem
ented
using a threshold based on the percentage change of the advertised res
ources
to avoid excessive consumption of link bandwidth and computational res
ources,
and to avoid instability in the TED.</t>
</list></t>
<t><list style="symbols">
<t>In a TE system, it is also desirable for the routing subsystem to
make the load splitting ratio among multiple paths (with equal cost
or different cost) configurable. This capability gives network
administrators more flexibility in the control of traffic administrators more flexibility in the control of traffic
distribution across the network. It can be very useful for distribution across the network. It can be very useful for
avoiding/relieving congestion in certain situations. Examples can be avoiding/relieving congestion in certain situations. Examples can
found in <xref target="XIAO"/> and <xref target="I-D.ietf-bess-evpn-un be found in <xref target="XIAO" format="default"/> and <xref
equal-lb" />.</t> target="I-D.ietf-bess-evpn-unequal-lb" format="default"/>.</li>
<li>The routing system should also have the capability to control
<t>The routing system should also have the capability to control the the routes of subsets of traffic without affecting the routes of
routes of subsets of traffic without affecting the routes of other other traffic if sufficient resources exist for this purpose. This
traffic if sufficient resources exist for this purpose. This
capability allows a more refined control over the distribution of capability allows a more refined control over the distribution of
traffic across the network. For example, the ability to move traffic traffic across the network. For example, the ability to move
away from its original path to another path (without affecting other traffic away from its original path to another path (without
traffic paths) allows the traffic to be moved from resource-poor netwo affecting other traffic paths) allows the traffic to be moved from
rk resource-poor network segments to resource-rich segments.
segments to resource-rich segments. Path oriented technologies such a Path-oriented technologies, such as MPLS-TE, inherently support this
s capability as discussed in <xref target="AWD2"
MPLS-TE inherently support this capability as discussed in <xref targe format="default"/>.</li>
t="AWD2"/>.</t> <li>Additionally, the routing subsystem should be able to select
<t>Additionally, the routing subsystem should be able to select
different paths for different classes of traffic (or for different different paths for different classes of traffic (or for different
traffic behavior aggregates) if the network supports multiple classes traffic behavior aggregates) if the network supports multiple
of service (different behavior aggregates).</t> classes of service (different behavior aggregates).</li>
</list></t> </ul>
</section>
</section> <section anchor="MAPREC" numbered="true" toc="default">
<name>Traffic Mapping Recommendations</name>
<section anchor="MAPREC" title="Traffic Mapping Recommendations"> <t>Traffic mapping is the assignment of traffic workload onto
(pre-established) paths to meet certain requirements. Thus, while
<t>Traffic mapping is the assignment of traffic workload onto constraint-based routing deals with path selection, traffic mapping
(pre-established) paths to meet certain requirements. Thus, while deals with the assignment of traffic to established paths that may
constraint-based routing deals with path selection, traffic mapping have been generated by constraint-based routing or by some other
deals with the assignment of traffic to established paths which may means. Traffic mapping can be performed by time-dependent or
have been generated by constraint-based routing or by some other state-dependent mechanisms, as described in <xref target="TIME"
means. Traffic mapping can be performed by time-dependent or state- format="default"/>.</t>
dependent mechanisms, as described in <xref target="TIME"/>.</t> <t>Two important aspects of the traffic mapping function are the
ability to establish multiple paths between an originating node and a
<t>An important aspect of the traffic mapping function is the ability to destination node and the capability to distribute the traffic across
establish multiple paths between an originating node and a destination those paths according to configured policies. A precondition for this
node, and the capability to distribute the traffic between the two scheme is the existence of flexible mechanisms to partition traffic
nodes across the paths according to configured policies. A precondition and then assign the traffic partitions onto the parallel paths
for this scheme is the existence of flexible mechanisms to partition (described as "parallel traffic trunks" in <xref target="RFC2702"
traffic and then assign the traffic partitions onto the parallel paths format="default"/>). When traffic is assigned to multiple parallel
(described as "parallel traffic trunks" in <xref target="RFC2702"/>). paths, it is recommended that special care should be taken to ensure
When traffic is assigned to multiple parallel paths, it is recommended proper ordering of packets belonging to the same application (or
that special care should be taken to ensure proper ordering of packets traffic flow) at the destination node of the parallel paths.</t>
belonging to the same application (or traffic flow) at the destination <t>Mechanisms that perform the traffic mapping functions should aim to
node of the parallel paths.</t> map the traffic onto the network infrastructure to minimize
congestion. If the total traffic load cannot be accommodated, or if
<t>Mechanisms that perform the traffic mapping functions should aim to map the routing and mapping functions cannot react fast enough to changing
the traffic onto the network infrastructure to minimize congestion. If traffic conditions, then a traffic mapping system may use short
the total traffic load cannot be accommodated, or if the routing and timescale congestion control mechanisms (such as queue management,
mapping functions cannot react fast enough to changing traffic scheduling, etc.) to mitigate congestion. Thus, mechanisms that
conditions, then a traffic mapping system may use short timescale perform the traffic mapping functions complement existing congestion
congestion control mechanisms (such as queue management, scheduling, control mechanisms. In an operational network, traffic should be
etc.) to mitigate congestion. Thus, mechanisms that perform the traffic mapped onto the infrastructure such that intra-class and inter-class
mapping functions complement existing congestion control mechanisms. In resource contention are minimized (see <xref target="BG"
an operational network, traffic should be mapped onto the infrastructure format="default"/>).</t>
such that intra-class and inter-class resource contention are minimized <t>When traffic mapping techniques that depend on dynamic state
(see <xref target="BG" />).</t> feedback (e.g., MPLS Adaptive Traffic Engineering (MATE) <xref target="M
ATE" format="default"/> and
<t>When traffic mapping techniques that depend on dynamic state feedback suchlike) are used, special care must be taken to guarantee network
(e.g., MATE <xref target="MATE" /> and suchlike) are used, special care stability.</t>
must be taken to guarantee network stability.</t> </section>
<section anchor="MSRREC" numbered="true" toc="default">
</section> <name>Measurement Recommendations</name>
<t>The importance of measurement in TE has been discussed throughout
<section anchor="MSRREC" title="Measurement Recommendations"> this document. A TE system should include mechanisms to measure and
collect statistics from the network to support the TE function.
<t>The importance of measurement in TE has been discussed Additional capabilities may be needed to help in the analysis of the
throughout this document. A TE system should include mechanisms to statistics. The actions of these mechanisms should not adversely
measure and collect statistics from the network to support the TE affect the accuracy and integrity of the statistics collected. The
function. Additional capabilities may be needed to help in the analysis mechanisms for statistical data acquisition should also be able to
of the statistics. The actions of these mechanisms should not adversely scale as the network evolves.</t>
affect the accuracy and integrity of the statistics collected. The <t>Traffic statistics may be classified according to long-term or
mechanisms for statistical data acquisition should also be able to scale short-term timescales. Long-term traffic statistics are very useful
as the network evolves.</t> for traffic engineering. Long-term traffic statistics may
periodically record network workload (such as hourly, daily, and
<t>Traffic statistics may be classified according to long-term or short-term weekly variations in traffic profiles) as well as traffic trends.
timescales. Long-term traffic statistics are very useful for traffic Aspects of the traffic statistics may also describe class of service
engineering. Long-term traffic statistics may periodically record networ characteristics for a network supporting multiple classes of service.
k Analysis of the long-term traffic statistics may yield other
workload (such as hourly, daily, and weekly variations in traffic profile information such as busy-hour characteristics, traffic growth
s) as patterns, persistent congestion problems, hotspots, and imbalances in
well as traffic trends. Aspects of the traffic statistics may also descr link utilization caused by routing anomalies.</t>
ibe <t>A mechanism for constructing traffic matrices for both long-term
class of service characteristics for a network supporting multiple classe and short-term traffic statistics should be in place. In
s of multi-service IP networks, the traffic matrices may be constructed for
service. Analysis of the long-term traffic statistics may yield other in different service classes. Each element of a traffic matrix
formation represents a statistic about the traffic flow between a pair of
such as busy-hour characteristics, traffic growth patterns, persistent co abstract nodes. An abstract node may represent a router, a collection
ngestion of routers, or a site in a VPN.</t>
problems, hot-spot, and imbalances in link utilization caused by routing <t>Traffic statistics should provide reasonable and reliable
anomalies.</t> indicators of the current state of the network on the short-term
scale. Some short-term traffic statistics may reflect link
<t>A mechanism for constructing traffic matrices for both long-term and shor utilization and link congestion status. Examples of congestion
t-term indicators include excessive packet delay, packet loss, and high
traffic statistics should be in place. In multi-service IP networks, the resource utilization. Examples of mechanisms for distributing this
traffic kind of information include SNMP, probing tools, FTP, IGP link-state
matrices may be constructed for different service classes. Each element advertisements, NETCONF/RESTCONF, etc.</t>
of a </section>
traffic matrix represents a statistic about the traffic flow between a pa <section anchor="POLICE" numbered="true" toc="default">
ir of <name>Policing, Planning, and Access Control</name>
abstract nodes. An abstract node may represent a router, a collection of <t>The recommendations in Sections <xref target="ROUTEREC" format="count
routers, er"/>
or a site in a VPN.</t> and <xref target="MAPREC" format="counter"/> may be sub-optimal or
even ineffective if the amount of traffic flowing on a route or path
<t>Traffic statistics should provide reasonable and reliable indicators of t exceeds the capacity of the resource on that route or path. Several
he current approaches can be used to increase the performance of TE systems:</t>
state of the network on the short-term scale. Some short term traffic st <ul spacing="normal">
atistics <li>The fundamental approach is some form of planning where traffic
may reflect link utilization and link congestion status. Examples of con is steered onto paths so that it is distributed across the available
gestion resources. This planning may be centralized or distributed and
indicators include excessive packet delay, packet loss, and high resource must be aware of the planned traffic volumes and available
utilization. resources. However, this approach is only of value if the traffic
Examples of mechanisms for distributing this kind of information include that is presented conforms to the planned traffic volumes.</li>
SNMP, probing <li>Traffic flows may be policed at the edges of a network. This is
tools, FTP, IGP link state advertisements, and NETCONF/RESTCONF, etc.</t> a simple way to ensure that the actual traffic volumes are
consistent with the planned volumes. Some form of measurement (see
</section> <xref target="MSRREC" format="default"/>) is used to determine the
rate of arrival of traffic, and excess traffic could be discarded.
<section anchor="POLICE" title="Policing, Planning, and Access Control"> Alternatively, excess traffic could be forwarded as best-effort
within the network. However, this approach is only completely
<t>The recommendations in <xref target="ROUTEREC" /> and <xref target="MAPRE effective if the planning is stringent and network-wide and if a
C" /> may be harsh approach is taken to disposing of excess traffic.</li>
sub-optimal or even ineffective if the amount of traffic flowing on a rou <li>Resource-based admission control is the process whereby network
te or path nodes decide whether to grant access to resources. The basis for
exceeds the capacity of the resource on that route or path. Several appr the decision on a packet-by-packet basis is the determination of the
oaches can be flow to which the packet belongs. This information is combined with
used to increase the performance of TE systems. policy instructions that have been locally configured or
installed through the management or control planes. The end result
<list style="symbols"> is that a packet may be allowed to access (or use) specific
resources on the node if, and only if, the flow to which the packet
<t>The fundamental approach is some form of planning where traffic is belongs conforms to the policy.</li>
steered onto paths so that it is </ul>
distributed across the available resources. This planning may be c <t>Combining some elements of all three of these measures is advisable
entralized or distributed, and to achieve a better TE system.</t>
must be aware of the planned traffic volumes and available resource </section>
s. However, this approach is <section anchor="SURVIVE" numbered="true" toc="default">
only of value if the traffic that is presented conforms to the plan <name>Network Survivability</name>
ned traffic volumes.</t> <t>Network survivability refers to the capability of a network to
maintain service continuity in the presence of faults. This can be
<t>Traffic flows may be policed at the edges of a network. This is a accomplished by promptly recovering from network impairments and
simple way to ensure that the maintaining the required QoS for existing services after recovery.
actual traffic volumes are consistent with the planned volumes. So Survivability is an issue of great concern within the Internet
me form of measurement (see community due to the demand to carry mission-critical traffic,
<xref target="MSRREC" />) is used to determine the rate of arrival real-time traffic, and other high-priority traffic over the Internet.
of traffic, and excess traffic Survivability can be addressed at the device level by developing
could be discarded. Alternatively, excess traffic could be forward network elements that are more reliable and at the network
ed as best-effort within the level by incorporating redundancy into the architecture, design, and
network. However, this approach is only completely effective if th operation of networks. It is recommended that a philosophy of
e planning is stringent and robustness and survivability should be adopted in the architecture,
network-wide, and if a harsh approach is taken to disposing of exce design, and operation of TE used to control IP networks (especially
ss traffic.</t> public IP networks). Because different contexts may demand different
levels of survivability, the mechanisms developed to support network
<t>Resource-based admission control is the process whereby network nod survivability should be flexible so that they can be tailored to
es decide whether to grant access different needs.
to resources. The basis for the decision on a packet-by-packet bas A number of tools and techniques have been developed
is is the determination of the flow to enable network survivability, including MPLS Fast Reroute <xref
to which the packet belongs. This information is combined with pol target="RFC4090" format="default"/>, Topology Independent Loop-free
icy instructions that have been Alternate Fast Reroute for Segment Routing <xref
locally configured, or installed through the management or control target="I-D.ietf-rtgwg-segment-routing-ti-lfa" format="default"/>,
planes. The end result is that RSVP-TE Extensions in Support of End-to-End GMPLS Recovery <xref
a packet may be allowed to access (or use) specific resources on th target="RFC4872" format="default"/>, and GMPLS Segment Recovery <xref
e node if and only if the flow to target="RFC4873" format="default"/>.</t>
which the packet belongs conforms to the policy.</t> <t>The impact of service outages varies significantly for different
service classes depending on the duration of the outage, which can vary
</list></t> from milliseconds (with minor service impact) to seconds (with
possible call drops for IP telephony and session timeouts for
<t>Combining some element of all three of these measures is advisable to ach connection-oriented transactions) to minutes and hours (with
ieve a better potentially considerable social and business impact). Outages of
TE system.</t> different durations have different impacts depending on the nature of
the traffic flows that are interrupted.</t>
</section> <t>Failure protection and restoration capabilities are available in
multiple layers as network technologies have continued to evolve.
<section anchor="SURVIVE" title="Network Survivability"> Optical networks are capable of providing dynamic ring and mesh
restoration functionality at the wavelength level. At the SONET/SDH
<t>Network survivability refers to the capability of a network to maintain s layer, survivability capability is provided with Automatic Protection
ervice Switching (APS) as well as self-healing ring and mesh architectures.
continuity in the presence of faults. This can be accomplished by prompt Similar functionality is provided by Layer 2 technologies such as
ly recovering Ethernet.</t>
from network impairments and maintaining the required QoS for existing se <t>Rerouting is used at the IP layer to restore service following link
rvices after and node outages. Rerouting at the IP layer occurs after a period of
recovery. Survivability is an issue of great concern within the Internet routing convergence, which may require seconds to minutes to complete.
community due Path-oriented technologies such as MPLS <xref target="RFC3469"
to the demand to carry mission-critical traffic, real-time traffic, and o format="default"/> can be used to enhance the survivability of IP
ther high networks in a potentially cost-effective manner.</t>
priority traffic over the Internet. Survivability can be addressed at th <t>An important aspect of multi-layer survivability is that
e device level technologies at different layers may provide protection and
by developing network elements that are more reliable; and at the network restoration capabilities at different granularities in terms of
level by timescales and at different bandwidth granularities (from the level of
incorporating redundancy into the architecture, design, and operation of packets to that of wavelengths). Protection and restoration
networks. It capabilities can also be sensitive to different service classes and
is recommended that a philosophy of robustness and survivability should b different network utility models. Coordinating different protection
e adopted in the and restoration capabilities across multiple layers in a cohesive
architecture, design, and operation of TE used to control IP networks manner to ensure network survivability is maintained at reasonable
(especially public IP networks). Because different contexts may demand d cost is a challenging task. Protection and restoration coordination
ifferent levels across layers may not always be feasible, because networks at different
of survivability, the mechanisms developed to support network survivabili layers may belong to different administrative domains.</t>
ty should be <t>Some of the general
flexible so that they can be tailored to different needs. A number of to recommendations for protection and restoration coordination are as follo
ols and techniques ws:</t>
have been developed to enable network survivability including MPLS Fast R <ul spacing="normal">
eroute <li>Protection and restoration capabilities from different layers
<xref target="RFC4090" />, Topology Independent Loop-free Alternate Fast should be coordinated to provide network survivability in a flexible
Re-route for and cost-effective manner. Avoiding duplication of functions in
Segment Routing <xref target="I-D.ietf-rtgwg-segment-routing-ti-lfa" /> R different layers is one way to achieve the coordination. Escalation
SVP-TE Extensions of alarms and other fault indicators from lower to higher layers may
in Support of End-to-End GMPLS Recovery <xref target="RFC4872" />, and GM also be performed in a coordinated manner. The order of timing of
PLS Segment Recovery restoration triggers from different layers is another way to
<xref target="RFC4873" />.</t> coordinate multi-layer protection/restoration.</li>
<li>Network capacity reserved in one layer to provide protection and
<t>The impact of service outages varies significantly for different service restoration is not available to carry traffic in a higher layer: it
classes depending is not visible as spare capacity in the higher layer. Placing
on the duration of the outage which can vary from milliseconds (with mino protection/restoration functions in many layers may increase
r service impact) redundancy and robustness, but it can result in significant
to seconds (with possible call drops for IP telephony and session timeout inefficiencies in network resource utilization. Careful planning is
s for needed to balance the trade-off between the desire for survivability
connection-oriented transactions) to minutes and hours (with potentially and the optimal use of resources.</li>
considerable social and business <li>It is generally desirable to have protection and restoration
impact). Outages of different durations have different impacts depending schemes that are intrinsically bandwidth efficient.</li>
on the nature of <li>Failure notifications throughout the network should be timely
the traffic flows that are interrupted.</t> and reliable if they are to be acted on as triggers for effective
protection and restoration actions.</li>
<t>Failure protection and restoration capabilities are available in multiple <li>Alarms and other fault monitoring and reporting capabilities
layers as network should be provided at the right network layers so that the
technologies have continued to evolve. Optical networks are capable of p protection and restoration actions can be taken in those
roviding dynamic layers.</li>
ring and mesh restoration functionality at the wavelength level. At the </ul>
SONET/SDH layer <section anchor="SRVMPLS" numbered="true" toc="default">
survivability capability is provided with Automatic Protection Switching <name>Survivability in MPLS-Based Networks</name>
(APS) as well as <t>Because MPLS is path-oriented, it has the potential to provide
self-healing ring and mesh architectures. Similar functionality is provi faster and more predictable protection and restoration capabilities
ded by layer 2 than conventional hop-by-hop routed IP systems. Protection types
technologies such as Ethernet.</t> for MPLS networks can be divided into four categories:</t>
<dl newline="false" spacing="normal">
<t>Rerouting is used at the IP layer to restore service following link and n <dt>Link Protection:</dt><dd>
ode outages. The objective of link protection is to protect an LSP from the
Rerouting at the IP layer occurs after a period of routing convergence wh failure of a given link. Under link protection, a protection or
ich may require backup LSP (the secondary LSP) follows a path that is disjoint
seconds to minutes to complete. Path-oriented technologies such as MPLS from the path of the working or operational LSP (the primary LSP)
(<xref target="RFC3469"/>) can be used to enhance the survivability of IP at the particular link where link protection is required. When
networks in a the protected link fails, traffic on the working LSP is switched
potentially cost-effective manner.</t> to the protection LSP at the headend of the failed link. As a
local repair method, link protection can be fast. This form of
<t>An important aspect of multi-layer survivability is that technologies at protection may be most appropriate in situations where some
different layers may network elements along a given path are known to be less reliable
provide protection and restoration capabilities at different granularitie than others.</dd>
s in terms of time <dt>Node Protection:</dt><dd>
scales and at different bandwidth granularity (from the level of packets The objective of node protection is to protect an LSP from the
to that of wavelengths). failure of a given node. Under node protection, the secondary LSP
Protection and restoration capabilities can also be sensitive to differen follows a path that is disjoint from the path of the primary LSP
t service classes at the particular node where node protection is required. The
and different network utility models. Coordinating different protection secondary LSP is also disjoint from the primary LSP at all links
and restoration attached to the node to be protected. When the protected node
capabilities across multiple layers in a cohesive manner to ensure networ fails, traffic on the working LSP is switched over to the
k survivability protection LSP at the upstream LSR directly connected to the
is maintained at reasonable cost is a challenging task. Protection and r failed node. Node protection covers a slightly larger part of the
estoration network compared to link protection but is otherwise
coordination across layers may not always be feasible, because networks a fundamentally the same.</dd>
t different <dt>Path Protection:</dt><dd>
layers may belong to different administrative domains.</t> The goal of LSP path protection (or end-to-end protection) is
to protect an LSP from any failure along its routed path. Under
<t>The following paragraphs present some of the general recommendations for path protection, the path of the protection LSP is completely
protection and disjoint from the path of the working LSP. The advantage of path
restoration coordination.</t> protection is that the backup LSP protects the working LSP from
all possible link and node failures along the path, except for
<t><list style="symbols"> failures of ingress or egress LSR. Additionally, path protection
<t>Protection and restoration capabilities from different layers should may be more efficient in terms of resource usage than link or node
be coordinated to provide protection applied at every hop along the path. However, path
network survivability in a flexible and cost-effective manner. Avoi protection may be slower than link and node protection because the
ding duplication of functions fault notifications have to be propagated further.</dd>
in different layers is one way to achieve the coordination. Escalat <dt>Segment Protection:</dt><dd>
ion of alarms and other fault An MPLS domain may be partitioned into multiple subdomains
indicators from lower to higher layers may also be performed in a co (protection domains). Path protection is applied to the path of
ordinated manner. The order each LSP as it crosses the domain from its ingress to the domain
of timing of restoration triggers from different layers is another w to where it egresses the domain. In cases where an LSP traverses
ay to coordinate multi-layer multiple protection domains, a protection mechanism within a
protection/restoration.</t> domain only needs to protect the segment of the LSP that lies
<t>Network capacity reserved in one layer to provide protection and res within the domain. Segment protection will generally be faster
toration is not available to than end-to-end path protection because recovery generally occurs
carry traffic in a higher layer: it is not visible as spare capacity closer to the fault, and the notification doesn't have to propagate
in the higher layer. Placing as far.</dd>
protection/restoration functions in many layers may increase redunda </dl>
ncy and robustness, but it can <t>See <xref target="RFC3469" format="default"/> and <xref
result in significant inefficiencies in network resource utilization target="RFC6372" format="default"/> for a more comprehensive
. Careful planning is needed discussion of MPLS-based recovery.</t>
to balance the tradeoff between the desire for survivability and the </section>
optimal use of resources.</t> <section anchor="PROTECT" numbered="true" toc="default">
<t>It is generally desirable to have protection and restoration schemes <name>Protection Options</name>
that are intrinsically <t>Another issue to consider is the concept of protection options.
bandwidth efficient.</t> We use notation such as "m:n protection", where m is the number of
<t>Failure notifications throughout the network should be timely and re protection LSPs used to protect n working LSPs. In all cases
liable if they are to be acted except 1+1 protection, the resources associated with the protection
on as triggers for effective protection and restoration actions.</t> LSPs can be used to carry preemptable best-effort traffic when the
<t>Alarms and other fault monitoring and reporting capabilities should working LSP is functioning correctly.</t>
be provided at the right network <dl spacing="normal" newline="false">
layers so that the protection and restoration actions can be taken i <dt>1:1 protection:</dt>
n those layers.</t> <dd>One working LSP is protected/restored by one protection LSP.
</list></t> Traffic is sent only on the protected LSP until the
protection/restoration event switches the traffic to the
<section anchor="SRVMPLS" title="Survivability in MPLS Based Networks"> protection LSP.</dd>
<dt>1:n protection:</dt>
<t>Because MPLS is path-oriented, it has the potential to provide faster a <dd>One protection LSP is used to protect/restore n working LSPs.
nd more predictable protection Traffic is sent only on the n protected working LSPs until the
and restoration capabilities than conventional hop-by-hop routed IP sys protection/restoration event switches the traffic from one failed
tems. Protection types for MPLS LSP to the protection LSP. Only one failed LSP can be restored at
networks can be divided into four categories.</t> any time.</dd>
<dt>n:1 protection:</dt>
<t><list style="symbols"> <dd>One working LSP is protected/restored by n protection LSPs,
<t>Link Protection: The objective of link protection is to protect an possibly with load splitting across the protection LSPs. This may
LSP from the failure of a given link. be especially useful when it is not feasible to find one path for
Under link protection, a protection or backup LSP (the secondary L the backup that can satisfy the bandwidth requirement of the
SP) follows a path that is disjoint primary LSP.</dd>
from the path of the working or operational LSP (the primary LSP) <dt>1+1 protection:</dt>
at the particular link where link <dd>Traffic is sent concurrently on both the working LSP and a
protection is required. When the protected link fails, traffic on protection LSP. The egress LSR selects one of the two LSPs based
the working LSP is switched to the on local policy (usually based on traffic integrity). When a
protection LSP at the head-end of the failed link. As a local rep fault disrupts the traffic on one LSP, the egress switches to
air method, link protection can be receive traffic from the other LSP. This approach is expensive in
fast. This form of protection may be most appropriate in situatio how it consumes network but recovers from failures most
ns where some network elements along a rapidly.</dd>
given path are known to be less reliable than others.</t> </dl>
<t>Node Protection: The objective of node protection is to protect an </section>
LSP from the failure of a given node. </section>
Under node protection, the secondary LSP follows a path that is di <section anchor="ML" numbered="true" toc="default">
sjoint from the path of the primary LSP <name>Multi-Layer Traffic Engineering</name>
at the particular node where node protection is required. The sec <t>Networks are often implemented as layers. A layer relationship may
ondary LSP is also disjoint from the represent the interaction between technologies (for example, an IP
primary LSP at all links attached to the node to be protected. Wh network operated over an optical network) or the relationship between
en the protected node fails, traffic on different network operators (for example, a customer network operated
the working LSP is switched over to the protection LSP at the upst over a service provider's network). Note that a multi-layer network
ream LSR directly connected to the does not imply the use of multiple technologies, although some form of
failed node. Node protection covers a slightly larger part of the encapsulation is often applied.</t>
network compared to link protection, <t>Multi-layer traffic engineering presents a number of challenges
but is otherwise fundamentally the same.</t> associated with scalability and confidentiality. These issues are
<t>Path Protection: The goal of LSP path protection (or end-to-end pr addressed in <xref target="RFC7926" format="default"/>, which
otection) is to protect an LSP from any discusses the sharing of information between domains through policy
failure along its routed path. Under path protection, the path of filters, aggregation, abstraction, and virtualization. That document
the protection LSP is completely disjoint also discusses how existing protocols can support this scenario with
from the path of the working LSP. The advantage of path protectio special reference to BGP-LS (see <xref target="BGPLS"
n is that the backup LSP protects the format="default"/>).</t>
working LSP from all possible link and node failures along the pat <t>PCE (see <xref target="PCE" format="default"/>) is also a useful
h, except for failures of ingress or tool for multi-layer networks as described in <xref target="RFC6805"
egress LSR. Additionally, path protection may be more efficient i format="default"/>, <xref target="RFC8685" format="default"/>, and
n terms of resource usage than link or <xref target="RFC5623" format="default"/>. Signaling techniques for
node protection applied at every hop along the path. However, pat multi-layer TE are described in <xref target="RFC6107"
h protection may be slower than link and format="default"/>.</t>
node protection because the fault notifications have to be propaga <t>See also <xref target="SURVIVE" format="default"/> for examination
ted further.</t> of multi-layer network survivability.</t>
<t>Segment Protection: An MPLS domain may be partitioned into multipl </section>
e subdomains (protection domains). Path <section anchor="TEDIFFSRV" numbered="true" toc="default">
protection is applied to the path of each LSP as it crosses the do <name>Traffic Engineering in Diffserv Environments</name>
main from its ingress to the domain to <t>Increasing requirements to support multiple classes of traffic in
where it egresses the domain. In cases where an LSP traverses mul the Internet, such as best-effort and mission-critical data, call for
tiple protection domains, a protection IP networks to differentiate traffic according to some criteria and to
mechanism within a domain only needs to protect the segment of the give preferential treatment to certain types of traffic. Large
LSP that lies within the domain. Segment numbers of flows can be aggregated into a few behavior aggregates
protection will generally be faster than end-to-end path protectio based on some criteria based on common performance requirements in
n because recovery generally occurs terms of packet loss ratio, delay, and jitter or in terms of common
closer to the fault and the notification doesn&apos;t have to prop fields within the IP packet headers.</t>
agate as far.</t> <t>Differentiated Services (Diffserv) <xref target="RFC2475"
</list></t> format="default"/> can be used to ensure that SLAs defined to
differentiate between traffic flows are met. Classes of service can
<t>See <xref target="RFC3469" /> and <xref target="RFC6372" /> for a more be supported in a Diffserv environment by concatenating Per-Hop
comprehensive discussion of MPLS Behaviors (PHBs) along the routing path. A PHB is the forwarding
based recovery.</t> behavior that a packet receives at a Diffserv-compliant node, and it
can be configured at each router. PHBs are delivered using
buffer-management and packet-scheduling mechanisms and require that
the ingress nodes use traffic classification, marking, policing, and
shaping.</t>
<t>TE can complement Diffserv to improve utilization of network
resources. TE can be operated on an aggregated basis across all
service classes <xref target="RFC3270" format="default"/> or on a
per-service-class basis. The former is used to provide better
distribution of the traffic load over the network resources (see <xref
target="RFC3270" format="default"/> for detailed mechanisms to support
aggregate TE). The latter case is discussed below since it is
specific to the Diffserv environment, with so-called Diffserv-aware
traffic engineering <xref target="RFC4124" format="default"/>.</t>
<t>For some Diffserv networks, it may be desirable to control the
performance of some service classes by enforcing relationships between
the traffic workload contributed by each service class and the amount
of network resources allocated or provisioned for that service class.
Such relationships between demand and resource allocation can be
enforced using a combination of, for example:
</t>
<ul spacing="normal">
<li>TE mechanisms on a per-service-class basis that enforce the
relationship between the amount of traffic contributed by a given
service class and the resources allocated to that class.</li>
<li>Mechanisms that dynamically adjust the resources allocated to a
given service class to relate to the amount of traffic contributed
by that service class.</li>
</ul>
<t>It may also be desirable to limit the performance impact of
high-priority traffic on relatively low-priority traffic. This can be
achieved, for example, by controlling the percentage of high-priority
traffic that is routed through a given link. Another way to
accomplish this is to increase link capacities appropriately so that
lower-priority traffic can still enjoy adequate service quality. When
the ratio of traffic workload contributed by different service classes
varies significantly from router to router, it may not be enough to
rely on conventional IGP routing protocols or on TE mechanisms that
are not sensitive to different service classes. Instead, it may be
desirable to perform TE, especially routing control and mapping
functions, on a per-service-class basis. One way to accomplish this
in a domain that supports both MPLS and Diffserv is to define
class-specific LSPs and to map traffic from each class onto one or
more LSPs that correspond to that service class. An LSP corresponding
to a given service class can then be routed and protected/restored in
a class-dependent manner, according to specific policies.</t>
<t>Performing TE on a per-class basis may require per-class parameters
to be distributed. It is common to have some classes share some
aggregate constraints (e.g., maximum bandwidth requirement) without
enforcing the constraint on each individual class. These classes can
be grouped into class types, and per-class-type parameters can be
distributed to improve scalability. This also allows better bandwidth
sharing between classes in the same class type. A class type is a set
of classes that satisfy the following two conditions:</t>
<ul spacing="normal">
<li>Classes in the same class type have common aggregate
requirements to satisfy required performance levels.</li>
<li>There is no requirement to be enforced at the level of an
individual class in the class type. Note that it is, nevertheless,
still possible to implement some priority policies for classes in
the same class type to permit preferential access to the class type
bandwidth through the use of preemption priorities.</li>
</ul>
<t>See <xref target="RFC4124" format="default"/> for detailed requiremen
ts on Diffserv-aware TE.</t>
</section>
<section anchor="CONTROL" numbered="true" toc="default">
<name>Network Controllability</name>
<t>Offline and online (see <xref target="OFFON" format="default"/>) TE
considerations are of limited utility if the network cannot be
controlled effectively to implement the results of TE decisions and to
achieve the desired network performance objectives.</t>
<t>Capacity augmentation is a coarse-grained solution to TE issues.
However, it is simple, may be applied through creating parallel links
that form part of an ECMP scheme, and may be advantageous if bandwidth
is abundant and cheap. However, bandwidth is not always abundant and
cheap, and additional capacity might not always be the best solution.
Adjustments of administrative weights and other parameters associated
with routing protocols provide finer-grained control, but this
approach is difficult to use and imprecise because of the way the
routing protocols interact across the network.</t>
<t>Control mechanisms can be manual (e.g., static configuration),
partially automated (e.g., scripts), or fully automated (e.g.,
policy-based management systems). Automated mechanisms are
particularly useful in large-scale networks. Multi-vendor
interoperability can be facilitated by standardized management tools
(e.g., YANG models) to support the control functions required to
address TE objectives.</t>
<t>Network control functions should be secure, reliable, and stable as
these are often needed to operate correctly in times of network
impairments (e.g., during network congestion or attacks).</t>
</section>
</section> </section>
<section anchor="INTER" numbered="true" toc="default">
<section anchor="PROTECT" title="Protection Options"> <name>Inter-Domain Considerations</name>
<t>Inter-domain TE is concerned with performance optimization for
<t>Another issue to consider is the concept of protection options. We use traffic that originates in one administrative domain and terminates in a
notation such as "m:n protection", where m is different one.</t>
the number of protection LSPs used to protect n working LSPs. In all c <t>BGP <xref target="RFC4271" format="default"/> is the standard
ases except 1+1 protection, the resources exterior gateway protocol used to exchange routing information between
associated with the protection LSPs can be used to carry preemptable be ASes in the Internet. BGP includes a decision
st-effort traffic when the working LSP is process that calculates the preference for routes to a given destination
functioning correctly.</t> network. There are two fundamental aspects to inter-domain TE using
BGP:</t>
<t><list style="symbols"> <dl newline="false" spacing="normal">
<t>1:1 protection: One working LSP is protected/restored by one prote <dt>Route Propagation:</dt>
ction LSP. Traffic is sent only on the protected <dd>Controlling the import and export of routes between ASes and
LSP until the protection/restoration event switches the traffic to controlling the redistribution of routes between BGP and other
the protection LSP.</t> protocols within an AS.</dd>
<t>1:n protection: One protection LSP is used to protect/restore n wo <dt>Best-path selection:</dt>
rking LSPs. Traffic is sent only on the n protected <dd>Selecting the best path when there are multiple candidate paths to
working LSPs until the protection/restoration event switches the t a given destination network.
raffic from one failed LSP to the protection LSP. This is performed by the BGP decision
Only one failed LSP can be restored at any time.</t> process, which selects the preferred exit points out of an AS toward spec
<t>n:1 protection: One working LSP is protected/restored by n protect ific
ion LSPs, possibly with load splitting across the destination networks by taking a number of different considerations into
protection LSPs. This may be especially useful when it is not fea account. The BGP path selection process can be influenced by
sible to find one path for the backup manipulating the attributes associated with the process, including
that can satisfy the bandwidth requirement of the primary LSP.</t> NEXT_HOP, LOCAL_PREF, AS_PATH, ORIGIN, MULTI_EXIT_DISC (MED), IGP
<t>1+1 protection: Traffic is sent concurrently on both the working L metric, etc.</dd>
SP and a protection LSP. The egress LSR selects </dl>
one of the two LSPs based on local policy (usually based on traffi <t>Most BGP implementations provide constructs that facilitate the
c integrity). When a fault disrupts the traffic implementation of complex BGP policies based on pre-configured logical
on one LSP, the egress switches to receive traffic from the other conditions. These can be used to control import and export of
LSP. This approach is expensive in how it consumes incoming and outgoing routes, control redistribution of routes
network but recovers from failures most rapidly.</t> between BGP and other protocols, and influence the selection of best
</list></t> paths by manipulating the attributes (either standardized or local to
the implementation) associated with the BGP decision process.</t>
<t>When considering inter-domain TE with BGP, note that the outbound
traffic exit point is controllable, whereas the interconnection point
where inbound traffic is received typically is not. Therefore, it is up
to each individual network to implement TE strategies that deal with the
efficient delivery of outbound traffic from its customers to its peering
points. The vast majority of TE policy is based on a "closest exit"
strategy, which offloads inter-domain traffic at the nearest outbound
peering point towards the destination AS. Most methods of manipulating
the point at which inbound traffic enters are either ineffective or
not accepted in the peering community.</t>
<t>Inter-domain TE with BGP is generally effective, but it is usually
applied in a trial-and-error fashion because a TE system usually only
has a view of the available network resources within one domain (an AS
in this case). A systematic approach for inter-domain TE requires
cooperation between the domains. Further, what may be considered a good
solution in one domain may not necessarily be a good solution in
another. Moreover, it is generally considered inadvisable for one
domain to permit a control process from another domain to influence the
routing and management of traffic in its network.</t>
<t>MPLS-TE tunnels (LSPs) can add a degree of flexibility in the
selection of exit points for inter-domain routing by applying the
concept of relative and absolute metrics. If BGP attributes are defined
such that the BGP decision process depends on IGP metrics to select exit
points for inter-domain traffic, then some inter-domain traffic destined
to a given peer network can be made to prefer a specific exit point by
establishing a TE tunnel between the router making the selection and the
peering point via a TE tunnel and assigning the TE tunnel a metric
that is smaller than the IGP cost to all other peering points. RSVP-TE
protocol extensions for inter-domain MPLS and GMPLS are described in
<xref target="RFC5151" format="default"/>.</t>
<t>Similarly to intra-domain TE, inter-domain TE is best accomplished
when a traffic matrix can be derived to depict the volume of traffic
from one AS to another.</t>
<t>Layer 4 multipath transport protocols are designed to move traffic
between domains and to allow some influence over the selection of the
paths. To be truly effective, these protocols would require visibility
of paths and network conditions in other domains, but that
information may not be available, might not be complete, and is not
necessarily trustworthy.</t>
</section> </section>
<section anchor="PRACTICE" numbered="true" toc="default">
<name>Overview of Contemporary TE Practices in Operational IP Networks</na
me>
<t>This section provides an overview of some TE practices in IP
networks. The focus is on aspects of control of the routing function in
operational contexts. The intent here is to provide an overview of the
commonly used practices; the discussion is not intended to be
exhaustive.</t>
<t>Service providers apply many of the TE mechanisms described in this
document to optimize the performance of their IP networks, although
others choose to not use any of them. These techniques include capacity
planning, including adding ECMP options, for long timescales; routing
control using IGP metrics and MPLS, as well as path planning and path
control using MPLS and SR for medium timescales; and
traffic management mechanisms for short timescales.</t>
<ul spacing="normal">
<li>Capacity planning is an important component of how a service
provider plans an effective IP network. These plans may take the
following aspects into account: location of any new links or nodes,
WECMP algorithms, existing and predicted traffic patterns, costs, link
capacity, topology, routing design, and survivability.</li>
<li>Performance optimization of operational networks is usually an
ongoing process in which traffic statistics, performance parameters,
and fault indicators are continually collected from the network. This
empirical data is analyzed and used to trigger TE mechanisms. Tools
that perform what-if analysis can also be used to assist the TE
process by reviewing scenarios before a new set of configurations are
implemented in the operational network.</li>
<li>Real-time intra-domain TE using the IGP is done by increasing the
OSPF or IS-IS metric of a congested link until enough traffic has been
diverted away from that link. This approach has some limitations as
discussed in <xref target="ROUTEREC" format="default"/>. Intra-domain
TE approaches <xref target="RR94" format="default"/> <xref
target="FT00" format="default"/> <xref target="FT01"
format="default"/> <xref target="WANG" format="default"/> take
traffic matrix, network topology, and network performance objectives
as input and produce link metrics and load-sharing ratios. These
processes open the possibility for intra-domain TE with IGP to be done
in a more systematic way.</li>
</ul>
</section> <t>Administrators of MPLS-TE networks specify and configure link
attributes and resource constraints such as maximum reservable bandwidth
<section anchor="ML" title="Multi-Layer Traffic Engineering"> and resource class attributes for the links in the domain. A link-state
IGP that supports TE extensions (IS-IS-TE or OSPF-TE) is used to
<t>Networks are often implemented as layers. A layer relationship may repr propagate information about network topology and link attributes to all
esent the interaction routers in the domain. Network administrators specify the LSPs that are
between technologies (for example, an IP network operated over an optica to originate at each router. For each LSP, the network administrator
l network), or the specifies the destination node and the attributes of the LSP that
relationship between different network operators (for example, a custome indicate the requirements that are to be satisfied during the path
r network operated selection process. The attributes may include an explicit path for the
over a service provider&apos;s network). Note that a multi-layer networ LSP to follow, or the originating router may use a local
k does not imply constraint-based routing process to compute the path of the LSP.
the use of multiple technologies, although some form of encapsulation is RSVP-TE is used as a signaling protocol to instantiate the LSPs. By
often applied.</t> assigning proper bandwidth values to links and LSPs, congestion caused
by uneven traffic distribution can be avoided or mitigated.</t>
<t>Multi-layer traffic engineering presents a number of challenges associat <t>The bandwidth attributes of an LSP relate to the bandwidth
ed with scalability requirements of traffic that flows through the LSP. The traffic
and confidentiality. These issues are addressed in <xref target="RFC792 attribute of an LSP can be modified to accommodate persistent shifts in
6" /> which discusses demand (traffic growth or reduction). If network congestion occurs due
the sharing of information between domains through policy filters, aggre to unexpected events, existing LSPs can be rerouted to alleviate the
gation, abstraction, situation, or the network administrator can configure new LSPs to divert
and virtualization. That document also discusses how existing protocols some traffic to alternative paths. The reservable bandwidth of the
can support this congested links can also be reduced to force some LSPs to be rerouted to
scenario with special reference to BGP-LS (see <xref target="BGPLS" />). other paths. A traffic matrix in an MPLS domain can also be estimated
</t> by monitoring the traffic on LSPs. Such traffic statistics can be used
for a variety of purposes including network planning and network
<t>PCE (see <xref target="PCE" />) is also a useful tool for multi-layer ne optimization.</t>
tworks as described <t>Network management and planning systems have evolved and assumed a
in <xref target="RFC6805" />, <xref target="RFC8685" />, and <xref targe lot of the responsibility for determining traffic paths in TE networks.
t="RFC5623" />. This allows a network-wide view of resources and facilitates
Signaling techniques for multi-layer TE are described in coordination of the use of resources for all traffic flows in the
<xref target="RFC6107" />.</t> network. Initial solutions using a PCE to perform path computation on
behalf of network routers have given way to an approach that follows the
<t>See also <xref target="SURVIVE" /> for examination of multi-layer networ SDN architecture. A stateful PCE is able to track all of the LSPs in
k survivability.</t> the network and can redistribute them to make better use of the
available resources. Such a PCE can form part of a network orchestrator
</section> that uses PCEP or some other configuration and management interface to
instruct the signaling protocol or directly program the routers.</t>
<section anchor="TEDIFFSRV" title="Traffic Engineering in Diffserv Environment <t>SR leverages a centralized TE controller and either an
s"> MPLS or IPv6 forwarding plane but does not need to use a signaling
protocol or management plane protocol to reserve resources in the
<t>Increasing requirements to support multiple classes of traffic in the Int routers. All resource reservation is logical within the controller and
ernet, such as is not distributed to the routers. Packets are steered through the
best effort and mission critical data, call for IP networks to differenti network using SR, and this may have configuration and
ate traffic operational scaling benefits.</t>
according to some criteria and to give preferential treatment to certain <t>As mentioned in <xref target="INTER" format="default"/>, there is
types of traffic. usually no direct control over the distribution of inbound traffic to a
Large numbers of flows can be aggregated into a few behavior aggregates b domain. Therefore, the main goal of inter-domain TE is to optimize the
ased on some distribution of outbound traffic between multiple inter-domain links.
criteria based on common performance requirements in terms of packet loss When operating a geographically widespread network (such as for a
ratio, delay, multi-national or global network provider), maintaining the ability to
and jitter, or in terms of common fields within the IP packet headers.</t operate the network in a regional fashion where desired, while
> continuing to take advantage of the benefits of a globally
interconnected network, also becomes an important objective.</t>
<t>Differentiated Services (Diffserv) <xref target="RFC2475"/> can be used t
o ensure that SLAs
defined to differentiate between traffic flows are met. Classes of servi
ce (CoS) can be
supported in a Diffserv environment by concatenating per-hop behaviors (P
HBs) along the
routing path. A PHB is the forwarding behavior that a packet receives at
a Diffserv-
compliant node, and it can be configured at each router. PHBs are delive
red using buffer
management and packet scheduling mechanisms and require that the ingress
nodes use traffic
classification, marking, policing, and shaping.</t>
<t>TE can complement Diffserv to improve utilization of network resources.
TE can be operated on an aggregated basis across all service classes
<xref target="RFC3270"/>, or on a per-service class basis. The former is
used to provide
better distribution of the traffic load over the network resources (see <
xref target="RFC3270"/>
for detailed mechanisms to support aggregate TE). The latter case is dis
cussed
below since it is specific to the Diffserv environment, with so called Di
ffserv-aware traffic
engineering <xref target="RFC4124"/>.</t>
<t>For some Diffserv networks, it may be desirable to control the performanc
e of some service
classes by enforcing relationships between the traffic workload contribut
ed by each service
class and the amount of network resources allocated or provisioned for th
at service class.
Such relationships between demand and resource allocation can be enforced
using a combination
of, for example:
<list style="symbols">
<t>TE mechanisms on a per service class basis that enforce the relation
ship between the amount
of traffic contributed by a given service class and the resources al
located to that class.</t>
<t>Mechanisms that dynamically adjust the resources allocated to a give
n service class to
relate to the amount of traffic contributed by that service class.</
t>
</list></t>
<t>It may also be desirable to limit the performance impact of high-priority
traffic on relatively
low-priority traffic. This can be achieved, for example, by controlling
the percentage of
high-priority traffic that is routed through a given link. Another way t
o accomplish this is to
increase link capacities appropriately so that lower-priority traffic can
still enjoy adequate
service quality. When the ratio of traffic workload contributed by diffe
rent service classes
varies significantly from router to router, it may not be enough to rely
on conventional IGP
routing protocols or on TE mechanisms that are not sensitive to different
service classes.
Instead, it may be desirable to perform TE, especially routing control an
d
mapping functions, on a per-service class basis. One way to accomplish t
his in a domain that
supports both MPLS and Diffserv is to define class-specific LSPs and to m
ap traffic from each
class onto one or more LSPs that correspond to that service class. An LS
P corresponding to a
given service class can then be routed and protected/restored in a class-
dependent manner,
according to specific policies.</t>
<t>Performing TE on a per-class basis may require per-class parameters to be
distributed. It is common to have some classes share some aggregate cons
traints (e.g., maximum
bandwidth requirement) without enforcing the constraint on each individua
l class. These classes
can be grouped into class-types, and per-class-type parameters can be dis
tributed to improve
scalability. This also allows better bandwidth sharing between classes i
n the same class-type.
A class-type is a set of classes that satisfy the following two condition
s:
<list style="symbols">
<t>Classes in the same class-type have common aggregate requirements to
satisfy required
performance levels.</t>
<t>There is no requirement to be enforced at the level of an individual
class in the class-type.
Note that it is, nevertheless, still possible to implement some prio
rity policies for classes
in the same class-type to permit preferential access to the class-ty
pe bandwidth through the
use of preemption priorities.</t>
</list></t>
<t>See <xref target="RFC4124"/> for detailed requirements on Diffserv-aware
TE.</t>
</section>
<section anchor="CONTROL" title="Network Controllability">
<t>Offline and online (see <xref target="OFFON" />) TE considerations are of
limited utility if the
network cannot be controlled effectively to implement the results of TE d
ecisions and to achieve
the desired network performance objectives.</t>
<t>Capacity augmentation is a coarse-grained solution to TE issues. However
, it is simple, may be applied
through creating parallel links that form part of an ECMP scheme, and may
be
advantageous if bandwidth is abundant and cheap. However, bandwidth is n
ot always abundant and cheap,
and additional capacity might not always be the best solution. Adjustmen
ts of administrative weights
and other parameters associated with routing protocols provide finer-grai
ned control, but this approach
is difficult to use and imprecise because of the way the routing protocol
s interactions occur across the
network.</t>
<t>Control mechanisms can be manual (e.g., static configuration), partially-
automated (e.g., scripts), or
fully-automated (e.g., policy based management systems). Automated mecha
nisms are particularly useful
in large-scale networks. Multi-vendor interoperability can be facilitate
d by standardized management
tools (e.g., YANG models) to support the control functions required to ad
dress TE objectives.</t>
<t>Network control functions should be secure, reliable, and stable as these
are often needed to operate
correctly in times of network impairments (e.g., during network congestio
n or attacks).</t>
</section>
</section>
<section anchor="INTER" title="Inter-Domain Considerations">
<t>Inter-domain TE is concerned with performance optimization for traffic that
originates in one administrative domain and terminates in a different one.<
/t>
<t>BGP <xref target="RFC4271"/> is the standard exterior gateway protocol used
to exchange
routing information between autonomous systems (ASes) in the Internet. BGP
includes a
decision process that calculates the preference for routes to a given
destination network. There are two fundamental aspects to inter-domain TE
using BGP:</t>
<t><list style="symbols">
<t>Route Propagation: Controlling the import and export of routes between
ASes, and
controlling the redistribution of routes between BGP and other protoco
ls within an AS.</t>
<t>Best-path selection: Selecting the best path when there are multiple c
andidate paths
to a given destination network. This is performed by the BGP decision
process, selecting
preferred exit points out of an AS towards specific destination networ
ks taking a
number of different considerations into account. The BGP path selecti
on process can
be influenced by manipulating the attributes associated with the proce
ss, including
NEXT_HOP, LOCAL_PREF, AS_PATH, ORIGIN, MULTI_EXIT_DISC (MED), IGP metr
ic, etc.</t>
</list></t>
<t>Most BGP implementations provide constructs that facilitate the implementat
ion of complex BGP
policies based on pre-configured logical conditions. These can be used to c
ontrol import and
export of incoming and outgoing routes, control redistribution of routes be
tween BGP and other
protocols, and influence the selection of best paths by manipulating the at
tributes (either
standardized, or local to the implementation) associated with the BGP decis
ion process.</t>
<t>When considering inter-domain TE with BGP, note that the outbound traffic e
xit point is controllable,
whereas the interconnection point where inbound traffic is received typical
ly is not. Therefore, it
is up to each individual network to implement TE strategies that deal with
the efficient delivery of
outbound traffic from its customers to its peering points. The vast majori
ty of TE policy is based
on a "closest exit" strategy, which offloads inter-domain traffic at the ne
arest outbound peering point
towards the destination AS. Most methods of manipulating the point at whic
h inbound traffic enters
are either ineffective, or not accepted in the peering community.</t>
<t>Inter-domain TE with BGP is generally effective, but it is usually applied
in a trial-and-error fashion
because a TE system usually only has a view of the available network resour
ces within one domain (an AS
in this case). A systematic approach for inter-domain TE requires cooperat
ion between the domains.
Further, what may be considered a good solution in one domain may not neces
sarily be a good solution in
another. Moreover, it is generally considered inadvisable for one domain t
o permit a control process from
another domain to influence the routing and management of traffic in its ne
twork.</t>
<t>MPLS TE-tunnels (LSPs) can add a degree of flexibility in the selection of
exit points for inter-domain routing
by applying the concept of relative and absolute metrics. If BGP attribute
s are defined such that the BGP
decision process depends on IGP metrics to select exit points for inter-dom
ain traffic, then some inter-domain
traffic destined to a given peer network can be made to prefer a specific e
xit point by establishing a TE-tunnel
between the router making the selection and the peering point via a TE-tunn
el and assigning the TE-tunnel a
metric which is smaller than the IGP cost to all other peering points. RSV
P-TE protocol extensions for inter-domain
MPLS and GMPLS are described in <xref target="RFC5151" />.</t>
<t>Similarly to intra-domain TE, inter-domain TE is best accomplished when a t
raffic matrix can be derived to
depict the volume of traffic from one AS to another.</t>
<t>Layer 4 multipath transport protocols are designed to move traffic between
domains and to allow some influence over the
selection of the paths. To be truly effective, these protocols would requi
re visibility of paths and network
conditions in other domains, and that information may not be available, mig
ht not be complete, and is not necessarily
trustworthy.</t>
</section>
<section anchor="PRACTICE" title="Overview of Contemporary TE Practices in Opera
tional IP Networks">
<t>This section provides an overview of some TE practices in IP networks. The
focus is on
aspects of control of the routing function in operational contexts. The in
tent here is to provide an
overview of the commonly used practices: the discussion is not intended to
be exhaustive.</t>
<t>Service providers apply many of the TE mechanisms described in this documen
t to optimize
the performance of their IP networks, although others choose to not use any
of them. These techniques
include capacity planning including adding ECMP options for long timescales
; routing control using IGP metrics and MPLS, as well as
path planning and path control using MPLS and Segment Routing for medium ti
mescales; and traffic management
mechanisms for short timescale.</t>
<list style="symbols">
<t>Capacity planning is an important component of how a service provider pl
ans an effective IP network.
These plans may take the following aspects into account: location of any
new links or nodes, WECMP
algorithms, existing and predicted traffic patterns, costs, link capacit
y, topology, routing design, and survivability.</t>
<t>Performance optimization of operational networks is usually an ongoing p
rocess in which traffic
statistics, performance parameters, and fault indicators are continually
collected from the network.
This empirical data is analyzed and used to trigger TE mechanisms. Tool
s that perform what-if analysis
can also be used to assist the TE process by reviewing scenarios before
a new set of configurations are
implemented in the operational network.</t>
<t>Real-time intra-domain TE using the IGP is done by increasing the OSPF o
r IS-IS metric of a congested
link until enough traffic has been diverted away from that link. This a
pproach has some limitations
as discussed in <xref target="ROUTEREC"/>. Intra-domain TE approaches (
<xref target="RR94"/>
<xref target="FT00"/> <xref target="FT01"/> <xref target="WANG"/>) take
traffic matrix, network topology,
and network performance objectives as input, and produce link metrics an
d load-sharing ratios. These
processes open the possibility for intra-domain TE with IGP to be done i
n a more systematic way.</t>
</list>
<t>Administrators of MPLS-TE networks specify and configure link attributes an
d resource constraints such
as maximum reservable bandwidth and resource class attributes for the links
in the domain. A link
state IGP that supports TE extensions (IS-IS-TE or OSPF-TE) is used to prop
agate information about
network topology and link attributes to all routers in the domain. Network
administrators specify
the LSPs that are to originate at each router. For each LSP, the network a
dministrator specifies the
destination node and the attributes of the LSP which indicate the requireme
nts that are to be satisfied
during the path selection process. The attributes may include an explicit
path for the LSP to follow, or
the originating router may use a local constraint-based routing process to
compute the path of the LSP. RSVP-TE
is used as a signaling protocol to instantiate the LSPs. By assigning prop
er bandwidth values to links
and LSPs, congestion caused by uneven traffic distribution can be avoided o
r mitigated.</t>
<t>The bandwidth attributes of an LSP relate to the bandwidth requirements of
traffic that flows through the
LSP. The traffic attribute of an LSP can be modified to accommodate persis
tent shifts in demand (traffic growth
or reduction). If network congestion occurs due to unexpected events, exis
ting LSPs can be rerouted to
alleviate the situation, or the network administrator can configure new LSP
s to divert some traffic to alternative
paths. The reservable bandwidth of the congested links can also be reduced
to force some LSPs to be rerouted
to other paths. A traffic matrix in an MPLS domain can also be estimated b
y monitoring the traffic on LSPs.
Such traffic statistics can be used for a variety of purposes including net
work planning and network
optimization.</t>
<t>Network management and planning systems have evolved and assumed a lot of t
he responsibility for determining
traffic paths in TE networks. This allows a network-wide view of resources
, and facilitates coordination of
the use of resources for all traffic flows in the network. Initial solutio
ns using a PCE to perform path
computation on behalf of network routers have given way to an approach that
follows the SDN architecture. A
stateful PCE is able to track all of the LSPs in the network and can redist
ribute them to make better use of
the available resources. Such a PCE can form part of a network orchestrato
r that uses PCEP or some other
configuration and management interface to instruct the signaling protocol o
r directly program the routers.</t>
<t>Segment Routing leverages a centralized TE controller and either an MPLS or
IPv6 forwarding plane, but does
not need to use a signaling protocol or management plane protocol to reserv
e resources in the routers. All
resource reservation is logical within the controller, and not distributed
to the routers. Packets are steered
through the network using Segment Routing, and this may have configuration
and operational scaling benefits.</t>
<t>As mentioned in <xref target="INTER"/>, there is usually no direct control
over the distribution of inbound
traffic to a domain. Therefore, the main goal of inter-domain TE is to opt
imize the distribution of outbound
traffic between multiple inter-domain links. When operating a geographical
ly widespread network (such as for a
multi-national or global network provider), maintaining the ability to oper
ate the network in a regional fashion
where desired, while continuing to take advantage of the benefits of a glob
ally interconnected network, also
becomes an important objective.</t>
<t>Inter-domain TE with BGP begins with the placement of multiple peering inte
rconnection points that are in close
proximity to traffic sources/destination, and offer lowest-cost paths acros
s the network between the peering
points and the sources/destinations. Some location-decision problems that
arise in association with
inter-domain routing are discussed in <xref target="AWD5"/>.</t>
<t>Once the locations of the peering interconnects have been determined and im
plemented, the network operator
decides how best to handle the routes advertised by the peer, as well as ho
w to propagate the peer&apos;s routes
within their network. One way to engineer outbound traffic flows in a netw
ork with many peering interconnects
is to create a hierarchy of peers. Generally, the shortest AS paths will b
e chosen to forward traffic but BGP
metrics can be used to prefer some peers and so favor particular paths. Pr
eferred peers are those peers attached
through peering interconnects with the most available capacity. Changes ma
y be needed, for example, to deal with
a "problem peer" who is difficult to work with on upgrades or is charging h
igh prices for connectivity to their
network. In that case, the peer may be given a reduced preference. This t
ype of change can affect a large amount
of traffic, and is only used after other methods have failed to provide the
desired results.</t>
<t>When there are multiple exit points toward a given peer, and only one of th
em is congested, it is not necessary
to shift traffic away from the peer entirely, but only from the one congest
ed connections. This can be
achieved by using passive IGP metrics, AS_PATH filtering, or prefix filteri
ng.</t>
</section>
<section anchor="SECURE" title="Security Considerations">
<t>In general, TE mechanisms are security-neutral, and this document does not
introduce new security issues.</t>
<t>Network security is, of course, an important issue, and TE mechanisms can h
ave benefits and drawbacks.</t>
<list style="symbols">
<t>TE may use tunnels which can slightly help protect traffic from inspecti
on and which, in some cases, can be
secured using encryption.</t>
<t>TE puts traffic onto predictable paths within the network that may make
it easier to find and attack.</t>
<t>TE often increases the complexity of operation and management of the net
work which may lead to errors that
compromise security.</t>
<t>TE enables traffic to be steered onto more secure links or to more secur
e parts of the network.</t>
<t>TE can be used to steer traffic through network nodes that are able to p
rovide additional security
functions.</t>
</list>
<t>The consequences of attacks on the control and management protocols used to
operate TE networks can be significant:
traffic can be hijacked to pass through specific nodes that perform inspect
ion, or even to be delivered to the
wrong place; traffic can be steered onto paths that deliver quality that is
below the desired quality; and, networks
can be congested or have resources on key links consumed. Thus, it is impo
rtant to use adequate protection mechanisms,
such as authentication, on all protocols used to deliver TE.</t>
<t>Certain aspects of a network may be deduced from the details of the TE path
s that are used. For example, the link
connectivity of the network, and the quality and load on individual links m
ay be inferred from knowing the paths of
traffic and the requirements they place on the network (for example, by see
ing the control messages or through path-
trace techniques). Such knowledge can be used to launch targeted attacks (
for example, taking down critical links)
or can reveal commercially sensitive information (for example, whether a ne
twork is close to capacity). Network
operators may, therefore, choose techniques that mask or hide information f
rom within the network.</t>
<t>External control interfaces that are introduced to provide additional contr
ol and management of TE systems (see
<xref target="TEapproach" />) provide flexibility to management and to cust
omers, but do so at the risk of
exposing the internals of a network to potentially malicious actors. The p
rotocols used at these interfaces
must be secured to protect against snooping and modification, and use of th
e interfaces must be authenticated.</t>
</section>
<section anchor="IANA" title="IANA Considerations">
<t>This draft makes no requests for IANA action.</t>
</section>
<section anchor="ACKN" title="Acknowledgments">
<t>Much of the text in this document is derived from RFC 3272. The editor and
contributors to this
document would like to express their gratitude to all involved in that work
.
Although the source text has been edited in the production of this document
, the
original authors should be considered as Contributors to this work. They w
ere:</t>
<figure><artwork><![CDATA[
Daniel O. Awduche
Movaz Networks
Angela Chiu
Celion Networks
Anwar Elwalid
Lucent Technologies
Indra Widjaja
Bell Labs, Lucent Technologies
XiPeng Xiao
Redback Networks
]]></artwork></figure>
<t>The acknowledgements in RFC3272 were as below. All people who helped
in the production of that document also need to be thanked for the
carry-over into this new document.</t>
<figure><artwork><![CDATA[
The authors would like to thank Jim Boyle for inputs on the
recommendations section, Francois Le Faucheur for inputs on
Diffserv aspects, Blaine Christian for inputs on measurement,
Gerald Ash for inputs on routing in telephone networks and for
text on event-dependent TE methods, Steven Wright for inputs
on network controllability, and Jonathan Aufderheide for
inputs on inter-domain TE with BGP. Special thanks to
Randy Bush for proposing the TE taxonomy based on "tactical versus
strategic" methods. The subsection describing an "Overview of
ITU Activities Related to Traffic Engineering" was adapted from
a contribution by Waisum Lai. Useful feedback and pointers to
relevant materials were provided by J. Noel Chiappa.
Additional comments were provided by Glenn Grotefeld during
the working last call process. Finally, the authors would like
to thank Ed Kern, the TEWG co-chair, for his comments and
support.
]]></artwork></figure>
<t>The early versions of this document were produced by the TEAS Working Group
&apos;s
RFC3272bis Design Team. The full list of members of this team is:</t>
<figure><artwork><![CDATA[
Acee Lindem
Adrian Farrel
Aijun Wang
Daniele Ceccarelli
Dieter Beller
Jeff Tantsura
Julien Meuric
Liu Hua
Loa Andersson
Luis Miguel Contreras
Martin Horneffer
Tarek Saad
Xufeng Liu
]]></artwork></figure>
<t>The production of this document includes a fix to the original text <t>Inter-domain TE with BGP begins with the placement of multiple
resulting from an Errata Report by Jean-Michel Grimaldi.</t> peering interconnection points that are in close proximity to traffic
sources/destinations and offer lowest-cost paths across the network
between the peering points and the sources/destinations. Some
location-decision problems that arise in association with inter-domain
routing are discussed in <xref target="AWD5" format="default"/>.</t>
<t>Once the locations of the peering interconnects have been determined
and implemented, the network operator decides how best to handle the
routes advertised by the peer, as well as how to propagate the peer's
routes within their network. One way to engineer outbound traffic flows
in a network with many peering interconnects is to create a hierarchy of
peers. Generally, the shortest AS paths will be chosen to forward
traffic, but BGP metrics can be used to prefer some peers and so favor
particular paths. Preferred peers are those peers attached through
peering interconnects with the most available capacity. Changes may be
needed, for example, to deal with a "problem peer" who is difficult to
work with on upgrades or is charging high prices for connectivity to
their network. In that case, the peer may be given a reduced
preference. This type of change can affect a large amount of traffic
and is only used after other methods have failed to provide the desired
results.</t>
<t>When there are multiple exit points toward a given peer, and only one
of them is congested, it is not necessary to shift traffic away from the
peer entirely, but only from the one congested connection. This can be
achieved by using passive IGP metrics, AS_PATH filtering, or prefix
filtering.</t>
</section>
<section anchor="SECURE" numbered="true" toc="default">
<name>Security Considerations</name>
<t>In general, TE mechanisms are security neutral, and this document
does not introduce new security issues.</t>
<t>Network security is, of course, an important issue, and TE mechanisms
can have benefits and drawbacks:</t>
<ul spacing="normal">
<li>TE may use tunnels that can slightly help protect traffic from
inspection and that, in some cases, can be secured using
encryption.</li>
<li>TE puts traffic onto predictable paths within the network that may
make it easier to find and attack.</li>
<li>TE often increases the complexity of operation and management of
the network, which may lead to errors that compromise security.</li>
<li>TE enables traffic to be steered onto more secure links or
to more secure parts of the network.</li>
<li>TE can be used to steer traffic through network nodes that are
able to provide additional security functions.</li>
</ul>
<t>The consequences of attacks on the control and management protocols
used to operate TE networks can be significant:</t>
<ul spacing="normal">
<li>Traffic can be hijacked
to pass through specific nodes that perform inspection or even to be
delivered to the wrong place.</li>
<li>Traffic can be steered onto paths that
deliver quality that is below the desired quality.</li>
<li>Networks can be
congested or have resources on key links consumed.</li>
</ul>
<t>Thus, it is
important to use adequate protection mechanisms, such as authentication,
on all protocols used to deliver TE.</t>
<t>Certain aspects of a network may be deduced from the details of the
TE paths that are used. For example, the link connectivity of the
network and the quality and load on individual links may be inferred
from knowing the paths of traffic and the requirements they place on the
network (for example, by seeing the control messages or through
path-trace techniques). Such knowledge can be used to launch targeted
attacks (for example, taking down critical links) or can reveal
commercially sensitive information (for example, whether a network is
close to capacity). Therefore, network operators may choose techniques
that mask or hide information from within the network.</t>
<t>External control interfaces that are introduced to provide additional
control and management of TE systems (see <xref target="TEapproach"
format="default"/>) provide flexibility to management and to customers,
but they do so at the risk of exposing the internals of a network to
potentially malicious actors. The protocols used at these interfaces
must be secured to protect against snooping and modification, and use of
the interfaces must be authenticated.</t>
</section>
<section anchor="IANA" numbered="true" toc="default">
<name>IANA Considerations</name>
<t>This document has no IANA actions.</t>
</section>
<t>The editor of this document would also like to thank Dhruv Dhody, Gyan Mish </middle>
ra, <back>
Joel Halpern, Dave Taht, John Scudder, Rich Salz, Behcet Sarikaya, Bob Bris
coe,
Erik Kline, Jim Guichard, Martin Duke, and Roman Danyliw, for review commen
ts.</t>
<t>This work is partially supported by the European Commission under <displayreference target="I-D.ietf-bess-evpn-unequal-lb" to="EVPN-UNEQUAL-LB"/>
Horizon 2020 grant agreement number 101015857 Secured autonomic <displayreference target="I-D.ietf-idr-performance-routing" to="PERFORMANCE-ROUT
traffic management for a Tera of SDN flows (Teraflow).</t> ING"/>
<displayreference target="I-D.ietf-idr-segment-routing-te-policy" to="SR-TE-POLI
CY"/>
<displayreference target="I-D.ietf-quic-multipath" to="QUIC-MULTIPATH"/>
<displayreference target="I-D.ietf-rtgwg-segment-routing-ti-lfa" to="SR-TI-LFA"/
>
<displayreference target="I-D.ietf-teas-enhanced-vpn" to="ENHANCED-VPN"/>
<displayreference target="I-D.ietf-tewg-qos-routing" to="TE-QoS-ROUTING"/>
<displayreference target="I-D.ietf-teas-ietf-network-slices" to="NETWORK-SLICES"
/>
<displayreference target="I-D.ietf-tsvwg-multipath-dccp" to="MULTIPATH-DCCP"/>
</section> <references>
<name>Informative References</name>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.079
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.110
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.110
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.220
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.233
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.238
6.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.247
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.247
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.259
7.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.267
8.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.270
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.272
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.275
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.296
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.299
8.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.303
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.308
6.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.312
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.316
8.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.317
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.319
8.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.320
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.327
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.327
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.346
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.347
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.363
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.394
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.409
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.412
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.420
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.427
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.434
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.446
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.459
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.465
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.487
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.487
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.487
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.515
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.525
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.530
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.532
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.533
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.535
7.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.539
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.544
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.547
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.547
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.554
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.555
7.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.555
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.562
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.566
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.567
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.569
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.610
7.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.611
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.624
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.637
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.637
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.660
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.680
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.701
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.714
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.728
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.739
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.742
6.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.747
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.749
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.755
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.756
7.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.766
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.767
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.768
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.955
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.792
6.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.792
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.795
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.803
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.803
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.804
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.805
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.818
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.823
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.825
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.827
9.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.828
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.828
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.829
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.840
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.845
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.857
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.857
1.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.865
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.866
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.868
4.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.868
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.879
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.880
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.889
6.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.893
8.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.895
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.897
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.900
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.902
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.904
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.911
3.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.925
6.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.926
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.929
8.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.931
5.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.933
2.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.935
0.xml"/>
<xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.943
9.xml"/>
<section anchor="CONTRIB" title="Contributors"> <reference anchor="Err309" quote-title="false" target="https://www.rfc-editor.or
g/errata/eid309">
<front>
<title>Erratum ID 309</title>
<author>
<organization>RFC Errata</organization>
</author>
</front>
<refcontent>RFC 3272</refcontent>
</reference>
<t>The following people contributed substantive text to this document:</t> <!-- [I-D.ietf-bess-evpn-unequal-lb] IESG state I-D Exists. Updated to lo ng version because missing editor role for Malhotra and contains extra initials for Lingala-->
<figure><artwork><![CDATA[ <reference anchor="I-D.ietf-bess-evpn-unequal-lb">
Gert Grammel <front>
EMail: ggrammel@juniper.net <title>Weighted Multi-Path Procedures for EVPN Multi-Homing</title>
<author initials="N." surname="Malhotra" fullname="Neeraj Malhotra" role="editor
">
<organization>Cisco Systems</organization>
</author>
<author initials="A." surname="Sajassi" fullname="Ali Sajassi">
<organization>Cisco Systems</organization>
</author>
<author initials="J." surname="Rabadan" fullname="Jorge Rabadan">
<organization>Nokia</organization>
</author>
<author initials="J." surname="Drake" fullname="John Drake">
<organization>Juniper</organization>
</author>
<author initials="A." surname="Lingala" fullname="Avinash Lingala">
<organization>ATT</organization>
</author>
<author initials="S." surname="Thoria" fullname="Samir Thoria">
<organization>Cisco Systems</organization>
</author>
<date month="December" day="7" year="2023"/>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-bess-evpn-unequal-lb-21"/>
</reference>
Loa Andersson <!-- [I-D.ietf-idr-performance-routing] IESG state Expired. Updated to long vers
EMail: loa@pi.nu ion because showing wrong date -->
Xufeng Liu <reference anchor="I-D.ietf-idr-performance-routing" target="https://datatracker
EMail: xufeng.liu.ietf@gmail.com .ietf.org/doc/html/draft-ietf-idr-performance-routing-03">
<front>
<title>Performance-based BGP Routing Mechanism</title>
<author initials="X." surname="Xu" fullname="Xiaohu Xu">
<organization>Alibaba, Inc</organization>
</author>
<author initials="S." surname="Hegde" fullname="Shraddha Hegde">
<organization>Juniper</organization>
</author>
<author initials="K." surname="Talaulikar" fullname="Ketan Talaulikar">
<organization>Cisco</organization>
</author>
<author initials="M." surname="Boucadair" fullname="Mohamed Boucadair">
<organization>France Telecom</organization>
</author>
<author initials="C." surname="Jacquenet" fullname="Christian Jacquenet">
<organization>France Telecom</organization>
</author>
<date month="December" day="22" year="2020"/>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-idr-performance-routing-03"/
>
</reference>
Lou Berger <!-- [I-D.ietf-idr-segment-routing-te-policy] IESG state AD is watching. Updated
EMail: lberger@labn.net to long version because missing editor role for Talaulikar -->
Jeff Tantsura <reference anchor="I-D.ietf-idr-segment-routing-te-policy">
EMail: jefftant.ietf@gmail.com <front>
<title>Advertising Segment Routing Policies in BGP</title>
<author initials="S." surname="Previdi" fullname="Stefano Previdi">
<organization>Huawei Technologies</organization>
</author>
<author initials="C." surname="Filsfils" fullname="Clarence Filsfils">
<organization>Cisco Systems</organization>
</author>
<author initials="K." surname="Talaulikar" fullname="Ketan Talaulikar" role="edi
tor">
<organization>Cisco Systems</organization>
</author>
<author initials="P." surname="Mattes" fullname="Paul Mattes">
<organization>Microsoft</organization>
</author>
<author initials="D." surname="Jain" fullname="Dhanendra Jain">
<organization>Google</organization>
</author>
<date month="October" day="23" year="2023"/>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-idr-segment-routing-te-polic
y-26"/>
</reference>
Daniel King <xi:include href="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.9502.xml"
EMail: daniel@olddog.co.uk />
Boris Hassanov <!-- [I-D.ietf-quic-multipath] IESG state I-D Exists. Updated to long version be
EMail: bhassanov@yandex-team.ru cause missing editor roles and fix name for Y. Ma-->
Kiran Makhijani <reference anchor="I-D.ietf-quic-multipath" target="https://datatracker.ietf.org
Email: kiranm@futurewei.com /doc/html/draft-ietf-quic-multipath-06">
<front>
<title>Multipath Extension for QUIC</title>
<author initials="Y." surname="Liu" fullname="Yanmei Liu" role="editor">
<organization>Alibaba Inc.</organization>
</author>
<author initials="Y." surname="Ma" fullname="Yunfei Ma" role="editor">
<organization>Uber Technologies Inc.</organization>
</author>
<author initials="Q." surname="De Coninck" fullname="Quentin De Coninck" role="e
ditor">
<organization>University of Mons (UMONS)</organization>
</author>
<author initials="O." surname="Bonaventure" fullname="Olivier Bonaventure">
<organization>UCLouvain and Tessares</organization>
</author>
<author initials="C." surname="Huitema" fullname="Christian Huitema">
<organization>Private Octopus Inc.</organization>
</author>
<author initials="M." surname="Kühlewind" fullname="Mirja Kühlewind" role="edito
r">
<organization>Ericsson</organization>
</author>
<date month="October" day="23" year="2023"/>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-quic-multipath-06"/>
</reference>
Dhruv Dhody <!-- [I-D.ietf-rtgwg-segment-routing-ti-lfa] IESG state I-D Exists -->
Email: dhruv.ietf@gmail.com
Mohamed Boucadair <xi:include href="https://datatracker.ietf.org/doc/bibxml3/reference.I-D.i
Email: mohamed.boucadair@orange.com etf-rtgwg-segment-routing-ti-lfa.xml"/>
]]></artwork></figure> <xi:include href="https://datatracker.ietf.org/doc/bibxml3/reference.I-D.i
etf-teas-enhanced-vpn.xml"/>
</section> <!-- [I-D.ietf-tewg-qos-routing] IESG state Expired (IESG: Dead). Updated to lon g version because showing wrong date -->
</middle> <reference anchor="I-D.ietf-tewg-qos-routing" target="https://datatracker.ietf.o
rg/doc/html/draft-ietf-tewg-qos-routing-04">
<front>
<title>Traffic Engineering &amp; QoS Methods for IP-, ATM-, &amp; Based Multiser
vice Networks</title>
<author initials="G." surname="Ash" fullname="Gerald Ash"> </author>
<date month="October" year="2001"/>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-tewg-qos-routing-04"/>
</reference>
<back> <!-- [I-D.ietf-teas-ietf-network-slices] IESG state IESG Evaluation::AD Followup . Updated to long version because missing editor roles -->
<references title='Informative References'> <reference anchor="I-D.ietf-teas-ietf-network-slices">
<front>
<title>A Framework for Network Slices in Networks Built from IETF Technologies</
title>
<author initials="A." surname="Farrel" fullname="Adrian Farrel" role="editor">
<organization>Old Dog Consulting</organization>
</author>
<author initials="J." surname="Drake" fullname="John Drake" role="editor">
<organization>Juniper Networks</organization>
</author>
<author initials="R." surname="Rokui" fullname="Reza Rokui">
<organization>Ciena</organization>
</author>
<author initials="S." surname="Homma" fullname="Shunsuke Homma">
<organization>NTT</organization>
</author>
<author initials="K." surname="Makhijani" fullname="Kiran Makhijani">
<organization>Futurewei</organization>
</author>
<author initials="L. M." surname="Contreras" fullname="Luis M. Contreras">
<organization>Telefonica</organization>
</author>
<author initials="J." surname="Tantsura" fullname="Jeff Tantsura">
<organization>Nvidia</organization>
</author>
<date month="September" day="14" year="2023"/>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-teas-ietf-network-slices-25"
/>
</reference>
&RFC0791; <!-- [I-D.ietf-tsvwg-multipath-dccp] IESG state I-D Exists. Updated to long vers
&RFC1102; ion because missing editor role -->
&RFC1104;
&RFC2205;
&RFC2330;
&RFC2386;
&RFC2474;
&RFC2475;
&RFC2597;
&RFC2678;
&RFC2702;
&RFC2722;
&RFC2753;
&RFC2961;
&RFC2998;
&RFC3031;
&RFC3086;
&RFC3124;
&RFC3168;
&RFC3175;
&RFC3198;
&RFC3209;
&RFC3270;
&RFC3272;
&RFC3469;
&RFC3473;
&RFC3630;
&RFC3945;
&RFC4090;
&RFC4124;
&RFC4203;
&RFC4271;
&RFC4340;
&RFC4461;
&RFC4594;
&RFC4655;
&RFC4872;
&RFC4873;
&RFC4875;
&RFC5151;
&RFC5250;
&RFC5305;
&RFC5329;
&RFC5331;
&RFC5357;
&RFC5394;
&RFC5440;
&RFC5470;
&RFC5472;
&RFC5541;
&RFC5557;
&RFC5559;
&RFC5623;
&RFC5664;
&RFC5671;
&RFC5693;
&RFC6107;
&RFC6119;
&RFC6241;
&RFC6372;
&RFC6374;
&RFC6601;
&RFC6805;
&RFC7011;
&RFC7149;
&RFC7285;
&RFC7390;
&RFC7426;
&RFC7471;
&RFC7491;
&RFC7551;
&RFC7567;
&RFC7665;
&RFC7679;
&RFC7680;
&RFC7752;
&RFC7926;
&RFC7923;
&RFC7950;
&RFC8033;
&RFC8034;
&RFC8040;
&RFC8051;
&RFC8189;
&RFC8231;
&RFC8259;
&RFC8279;
&RFC8281;
&RFC8283;
&RFC8290;
&RFC8402;
&RFC8453;
&RFC8570;
&RFC8571;
&RFC8655;
&RFC8664;
&RFC8684;
&RFC8685;
&RFC8795;
&RFC8803;
&RFC8896;
&RFC8938;
&RFC8955;
&RFC8972;
&RFC9000;
&RFC9023;
&RFC9040;
&RFC9113;
&RFC9256;
&RFC9262;
&RFC9298;
&RFC9315;
&RFC9332;
&RFC9350;
&RFC9439;
&I-D.ietf-bess-evpn-unequal-lb; <reference anchor="I-D.ietf-tsvwg-multipath-dccp">
&I-D.ietf-idr-performance-routing; <front>
&I-D.ietf-idr-segment-routing-te-policy; <title>DCCP Extensions for Multipath Operation with Multiple Addresses</title>
&I-D.ietf-lsr-ip-flexalgo; <author initials="M." surname="Amend" fullname="Markus Amend" role="editor">
&I-D.ietf-quic-multipath; <organization>Deutsche Telekom</organization>
&I-D.ietf-rtgwg-segment-routing-ti-lfa; </author>
&I-D.ietf-teas-enhanced-vpn; <author initials="A." surname="Brunstrom" fullname="Anna Brunstrom">
&I-D.ietf-tewg-qos-routing; <organization>Karlstad University</organization>
&I-D.ietf-teas-ietf-network-slices; </author>
&I-D.ietf-tsvwg-multipath-dccp; <author initials="A." surname="Kassler" fullname="Andreas Kassler">
<organization>Karlstad University</organization>
</author>
<author initials="V." surname="Rakocevic" fullname="Veselin Rakocevic">
<organization>City University of London</organization>
</author>
<author initials="S." surname="Johnson" fullname="Stephen Johnson">
<organization>BT</organization>
</author>
<date month="October" day="12" year="2023"/>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-tsvwg-multipath-dccp-11"/>
</reference>
<reference anchor="AFD03" target="https://dl.acm.org/doi/10.1145/956981.956985 <reference anchor="AFD03" target="https://dl.acm.org/doi/10.1145/956981.95
"> 6985">
<front> <front>
<title>Approximate fairness through differential dropping</title> <title>Approximate fairness through differential dropping</title>
<author initials="R." surname="Pan" fullname="Rong Pan"> <author initials="R." surname="Pan" fullname="Rong Pan">
<organization></organization> <organization/>
</author> </author>
<author initials="L." surname="Breslau" fullname="Lee Breslau"> <author initials="L." surname="Breslau" fullname="Lee Breslau">
<organization></organization> <organization/>
</author> </author>
<author initials="B." surname="Prabhakar" fullname="Balaji Prabhakar"> <author initials="B." surname="Prabhakar" fullname="Balaji Prabhakar">
<organization></organization> <organization/>
</author> </author>
<author initials="S." surname="Shenker" fullname="Scott Shenker"> <author initials="S." surname="Shenker" fullname="Scott Shenker">
<organization></organization> <organization/>
</author> </author>
<date year="2003"/> <date month="April" year="2003"/>
</front> </front>
<seriesInfo name="Article" value="ACM SIGCOMM Computer Communication Review, <refcontent>ACM SIGCOMM Computer Communication Review, Volume 33,
Volume 33, Issue 2, April 2003, pp 23-39"/> Issue 2, Pages 23-39</refcontent>
</reference> <seriesInfo name="DOI" value="10.1145/956981.956985"/>
</reference>
<reference anchor="AJ19" target="https://journalofbigdata.springeropen.com/tra <reference anchor="AJ19" target="https://journalofbigdata.springeropen.com
ck/pdf/10.1186/s40537-019-0176-5.pdf"> /track/pdf/10.1186/s40537-019-0176-5.pdf">
<front> <front>
<title>Data mining approach for predicting the daily Internet data <title>Data mining approach for predicting the daily Internet data
traffic of a smart university</title> traffic of a smart university</title>
<author initials="A." surname="Adekitan" fullname="A. Adekitan"> <author initials="A." surname="Adekitan" fullname="A. Adekitan">
<organization></organization> <organization/>
</author> </author>
<author initials="J." surname="Abolade" fullname="J. Abolade"> <author initials="J." surname="Abolade" fullname="J. Abolade">
<organization></organization> <organization/>
</author> </author>
<author initials="O." surname="Shobayo" fullname="O. Shobayo"> <author initials="O." surname="Shobayo" fullname="O. Shobayo">
<organization></organization> <organization/>
</author> </author>
<date year="1998"/> <date month="February" year="2019"/>
</front> </front>
<seriesInfo name="Article" value="Journal of Big Data, 2019, Volume 6, Numbe <refcontent>Journal of Big Data, Volume 6, Number 1, Page 1</refcontent>
r 1, Page 1"/> <seriesInfo name="DOI" value="10.1186/s40537-019-0176-5"/>
</reference> </reference>
<reference anchor="ATSSS" target="https://www.3gpp.org/ftp//Specs/archive/23_s <reference anchor="ATSSS" target="https://www.3gpp.org/ftp//Specs/archive/
eries/23.793/23793-g00.zip"> 23_series/23.793/23793-g00.zip">
<front> <front>
<title>Study on access traffic steering, switch and splitting support in t <title>Study on access traffic steering, switch and splitting
he 5G System (5GS) architecture</title> support in the 5G System (5GS) architecture</title>
<author > <author>
<organization></organization> <organization>3GPP</organization>
</author> </author>
<date year="2018" month="December"/> <date year="2018" month="December"/>
</front> </front>
<seriesInfo name="Specification" value="3GPP Technical Report 23.793 Release <refcontent>Release 16</refcontent>
16"/> <refcontent>3GPP TR 23.793</refcontent>
</reference> </reference>
<reference anchor="AWD2" target="https://ieeexplore.ieee.org/document/809383"> <reference anchor="AWD2" target="https://ieeexplore.ieee.org/document/8093
<front> 83">
<title>MPLS and Traffic Engineering in IP Networks</title> <front>
<author initials="D." surname="Awduche" fullname="Daniel Awduche"> <title>MPLS and traffic engineering in IP networks</title>
<organization></organization> <author initials="D." surname="Awduche" fullname="Daniel Awduche">
</author> <organization/>
<date year="1999" month="December"/> </author>
</front> <date year="1999" month="December"/>
<seriesInfo name="Article" value="IEEE Communications Magazine"/> </front>
</reference> <refcontent>IEEE Communications Magazine, Volume 37, Issue 12, Pages
42-47</refcontent>
<seriesInfo name="DOI" value="10.1109/35.809383"/>
</reference>
<reference anchor="AWD5" target="https://ieeexplore.ieee.org/document/998795"> <reference anchor="AWD5" target="https://ieeexplore.ieee.org/document/9987
<front> 95">
<title>An Approach to Optimal Peering Between Autonomous Systems in the In <front>
ternet</title> <title>An approach to optimal peering between autonomous systems in
<author initials="D." surname="Awduche" fullname="Daniel Awduche"> the Internet</title>
<organization></organization> <author initials="D." surname="Awduche" fullname="Daniel Awduche">
</author> <organization/>
<date year="1998" month="October"/> </author>
</front> <date year="1998" month="October"/>
<seriesInfo name="Paper" value="International Conference on Computer Communi </front>
cations and Networks (ICCCN'98)"/> <refcontent>Proceedings 7th International Conference on Computer
</reference> Communications and Networks (Cat. No. 98EX226)</refcontent>
<seriesInfo name="DOI" value="10.1109/ICCCN.1998.998795"/>
</reference>
<reference anchor="E.360.1" target="https://www.itu.int/rec/T-REC-E.360.1-2002 <reference anchor="E.360.1" target="https://www.itu.int/rec/T-REC-E.360.1-
05-I/en"> 200205-I/en">
<front> <front>
<title>E.360.1: Framework for QoS routing and related traffic engineering <title>Framework for QoS routing and related traffic
methods for IP-, ATM-, and TDM-based multiservice networks</title> engineering methods for IP-, ATM-, and TDM-based multiservice
<author> networks</title>
<organization>International Telecommunication Union - Telecommunication <author>
Standardization Sector</organization> <organization>ITU-T</organization>
</author> </author>
<date year="2002" month="May" day="16" /> <date year="2002" month="May"/>
</front> </front>
<seriesInfo name="Recommendation" value="ITU-T Recommendation E.360.1" /> <seriesInfo name="ITU-T Recommendation" value="E.360.1"/>
</reference> </reference>
<reference anchor="FLJA93" target="https://www.icir.org/floyd/papers/early.two <reference anchor="FLJA93" target="https://www.icir.org/floyd/papers/early
column.pdf"> .twocolumn.pdf">
<front> <front>
<title>Random Early Detection Gateways for Congestion Avoidance</title> <title>Random Early Detection Gateways for Congestion Avoidance</title
<author initials="S." surname="Floyd"> >
<organization></organization> <author initials="S." surname="Floyd">
</author> <organization/>
<author initials="V." surname="Jacobson"> </author>
<organization></organization> <author initials="V." surname="Jacobson">
</author> <organization/>
<date year="1993" month="November"/> </author>
</front> <date year="1993" month="August"/>
<seriesInfo name="Article" value="IEEE/ACM Transactions on Networking, Vol. </front>
1, p. 387-413"/> <refcontent>IEEE/ACM Transactions on Networking, Volume 1,
</reference> Issue 4, Pages 397-413</refcontent>
<seriesInfo name="DOI" value="10.1109/90.251892"/>
</reference>
<reference anchor="FT00" target="https://www.cs.cornell.edu/courses/cs619/2004 <reference anchor="FT00" target="https://www.cs.cornell.edu/courses/cs619/
fa/documents/ospf_opt.pdf"> 2004fa/documents/ospf_opt.pdf">
<front> <front>
<title>Internet Traffic Engineering by Optimizing OSPF Weights</title> <title>Internet Traffic Engineering by Optimizing OSPF Weights</title>
<author initials="B." surname="Fortz"> <author initials="B." surname="Fortz">
<organization></organization> <organization/>
</author> </author>
<author initials="M." surname="Thorup"> <author initials="M." surname="Thorup">
<organization></organization> <organization/>
</author> </author>
<date year="2000" month="March"/> <date year="2000" month="March"/>
</front> </front>
<seriesInfo name="Article" value="IEEE INFOCOM 2000"/> <refcontent>Proceedings IEEE INFOCOM 2000</refcontent>
</reference> <seriesInfo name="DOI" value="10.1109/INFCOM.2000.832225"/>
</reference>
<reference anchor="FT01" target="http://www.research.att.com/~mthorup/PAPERS/p <reference anchor="FT01" target="https://ieeexplore.ieee.org/document/1003
apers.html"> 042">
<front> <front>
<title>Optimizing OSPF/IS-IS Weights in a Changing World</title> <title>Optimizing OSPF/IS-IS Weights in a Changing World</title>
<author initials="B." surname="Fortz"> <author initials="B." surname="Fortz">
<organization></organization> <organization/>
</author> </author>
<author initials="M." surname="Thorup"> <author initials="M." surname="Thorup">
<organization></organization> <organization/>
</author> </author>
<date year="n.d."/> <date month="May" year="2002"/>
</front> </front>
</reference> <refcontent>IEEE Journal on Selected Areas in Communications</refcontent
>
<seriesInfo name='DOI' value='10.1109/JSAC.2002.1003042' />
</reference>
<reference anchor="GRPC" target="https://grpc.io"> <reference anchor="GRPC" target="https://grpc.io">
<front> <front>
<title>gPPC: A high performance, open source universal RPC framework</titl <title>gRPC: A high performance, open source universal RPC
e> framework</title>
<author> <author>
<organization>Cloud Native Computing Foundation</organization> <organization>gRPC Authors</organization>
</author> </author>
<date year="n.d."/> <date></date>
</front> </front>
</reference> </reference>
<reference anchor="KELLY"> <reference anchor="KELLY">
<front> <front>
<title>Notes on effective bandwidths. In Stochastic Networks: Theory and A <title>Notes on effective bandwidths</title>
pplications</title> <author initials="F." surname="Kelly">
<author initials="F." surname="Kelly"> </author>
</author> <date year="1996"/>
<date year="1996"/> </front>
</front> <refcontent>Oxford University Press</refcontent>
<seriesInfo name="Book" value="Oxford University Press"/> </reference>
</reference>
<reference anchor="MA" target="https://apps.dtic.mil/sti/pdfs/ADA352299.pdf"> <reference anchor="MA" target="https://apps.dtic.mil/sti/pdfs/ADA352299.pd
<front> f">
<title>Quality of Service Routing in Integrated Services Networks</title> <front>
<author initials="Q." surname="Ma"> <title>Quality-of-Service Routing in Integrated Services Networks</tit
<organization></organization> le>
</author> <author initials="Q." surname="Ma">
<date year="1998"/> <organization/>
</front> </author>
<seriesInfo name="Ph.D." value="PhD Dissertation, CMU-CS-98-138, CMU"/> <date month="January" year="1998"/>
</reference> </front>
<refcontent>Ph.D. Dissertation, Carnegie Mellon University,
CMU-CS-98-138</refcontent>
</reference>
<reference anchor="MATE" target="https://www.yumpu.com/en/document/view/351403 <reference anchor="MATE" target="https://www.yumpu.com/en/document/view/35
98/mate-mpls-adaptive-traffic-engineering-infocom-ieee-xplore/8"> 140398/mate-mpls-adaptive-traffic-engineering-infocom-ieee-xplore/8">
<front> <front>
<title>MATE - MPLS Adaptive Traffic Engineering</title> <title>MATE: MPLS Adaptive Traffic Engineering</title>
<author initials="A." surname="Elwalid"> <author initials="A." surname="Elwalid">
<organization></organization> <organization/>
</author> </author>
<author initials="C." surname="Jin"> <author initials="C." surname="Jin">
<organization></organization> <organization/>
</author> </author>
<author initials="S." surname="Low"> <author initials="S." surname="Low">
<organization></organization> <organization/>
</author> </author>
<author initials="I." surname="Widjaja"> <author initials="I." surname="Widjaja">
<organization></organization> <organization/>
</author> </author>
<date year="2001" month="April"/> <date year="2002" month="August"/>
</front> </front>
<seriesInfo name="Proceedings" value="INFOCOM'01"/> <refcontent>Proceedings IEEE INFOCOM 2001, Conference on Computer
</reference> Communications, Twentieth Annual Joint Conference of the IEEE Computer
and Communications Society (Cat. No. 01CH37213)</refcontent>
<seriesInfo name="DOI" value="10.1109/INFCOM.2001.916625"/>
</reference>
<reference anchor="MR99" target="https://ieeexplore.ieee.org/document/830281"> <reference anchor="MR99" target="https://ieeexplore.ieee.org/document/8302
<front> 81">
<title>A Case Study of Multiservice, Multipriority Traffic Engineering Des <front>
ign for Data Networks</title> <title>A case study of multiservice, multipriority traffic engineering
<author initials="D." surname="Mitra"> design for data networks</title>
<organization></organization> <author initials="D." surname="Mitra">
</author> <organization/>
<author initials="K.G." surname="Ramakrishnan"> </author>
<organization></organization> <author initials="K.G." surname="Ramakrishnan">
</author> <organization/>
<date year="1999" month="December"/> </author>
</front> <date year="1999" month="December"/>
<seriesInfo name="Proceedings" value="Globecom'99"/> </front>
</reference> <refcontent>Seamless Interconnection for Universal Services, Global
Telecommunications Conference, GLOBECOM'99,
(Cat. No. 99CH37042)</refcontent>
<seriesInfo name="DOI" value="10.1109/GLOCOM.1999.830281"/>
</reference>
<reference anchor="RR94" target="https://onlinelibrary.wiley.com/doi/abs/10.10 <reference anchor="RR94" target="https://onlinelibrary.wiley.com/doi/abs/1
02/bltj.2267"> 0.1002/bltj.2267">
<front> <front>
<title>Optimal Routing in Shortest Path Data Networks</title> <title>Optimal routing in shortest-path data networks</title>
<author initials="M.A." surname="Rodrigues"> <author initials="M." surname="Rodrigues">
<organization></organization> <organization/>
</author> </author>
<author initials="K.G." surname="Ramakrishnan"> <author initials="K.G." surname="Ramakrishnan">
<organization></organization> <organization/>
</author> </author>
<date year="1994"/> <date month="August" year="2002"/>
</front> </front>
<seriesInfo name="Proceedings" value="ITS'94, Rio de Janeiro, Brazil"/> <refcontent>Bell Labs Technical Journal, Volume 6, Issue 1, Pages
</reference> 117-138</refcontent>
<seriesInfo name="DOI" value="10.1002/bltj.2267"/>
</reference>
<reference anchor="SLDC98" target="https://ieeexplore.ieee.org/document/659666 <reference anchor="SLDC98" target="https://ieeexplore.ieee.org/document/65
"> 9666">
<front> <front>
<title>Design Considerations for Supporting TCP with Per-flow Queueing</ti <title>Design considerations for supporting TCP with per-flow queueing
tle> </title>
<author initials="B." surname="Suter"> <author initials="B." surname="Suter">
<organization></organization> <organization/>
</author> </author>
<author initials="T." surname="Lakshman"> <author initials="T.V." surname="Lakshman">
<organization></organization> <organization/>
</author> </author>
<author initials="D." surname="Stiliadis"> <author initials="D." surname="Stiliadis">
<organization></organization> <organization/>
</author> </author>
<author initials="A." surname="Choudhury"> <author initials="A.K." surname="Choudhury">
<organization></organization> <organization/>
</author> </author>
<date year="1998"/> <date month="April" year="1998"/>
</front> </front>
<seriesInfo name="Proceedings" value="INFOCOM'98, p. 299-306"/> <refcontent>Proceedings IEEE INFOCOM '98
</reference> </refcontent>
<seriesInfo name="DOI" value="10.1109/INFCOM.1998.659666"/>
</reference>
<reference anchor="WANG" target="https://ieeexplore.ieee.org/document/916782"> <reference anchor="WANG" target="https://ieeexplore.ieee.org/document/9167
<front> 82">
<title>Internet traffic engineering without full mesh overlaying</title> <front>
<author initials="Y." surname="Wang"> <title>Internet traffic engineering without full mesh overlaying</titl
<organization></organization> e>
</author> <author initials="Y." surname="Wang">
<author initials="Z." surname="Wang"> <organization/>
<organization></organization> </author>
</author> <author initials="Z." surname="Wang">
<author initials="L." surname="Zhang"> <organization/>
<organization></organization> </author>
</author> <author initials="L." surname="Zhang">
<date year="2001" month="April"/> <organization/>
</front> </author>
<seriesInfo name="Proceedings" value="INFOCOM'2001"/> <date year="2001" month="April"/>
</reference> </front>
<refcontent>Proceedings IEEE INFOCOM 2001
</refcontent>
<seriesInfo name="DOI" value="10.1109/INFCOM.2001.916782"/>
</reference>
<reference anchor="XIAO" target="https://courses.cs.washington.edu/courses/cse <reference anchor="XIAO" target="https://courses.cs.washington.edu/courses
561/02au/papers/xiao-mpls-net00.pdf"> /cse561/02au/papers/xiao-mpls-net00.pdf">
<front> <front>
<title>Traffic Engineering with MPLS in the Internet</title> <title>Traffic Engineering with MPLS in the Internet</title>
<author initials="X." surname="Xiao"> <author initials="X." surname="Xiao">
<organization></organization> <organization/>
</author> </author>
<author initials="A." surname="Hannan"> <author initials="A." surname="Hannan">
<organization></organization> <organization/>
</author> </author>
<author initials="B." surname="Bailey"> <author initials="B." surname="Bailey">
<organization></organization> <organization/>
</author> </author>
<author initials="L." surname="Ni"> <author initials="L." surname="Ni">
<organization></organization> <organization/>
</author> </author>
<date year="2000" month="March"/> <date year="2000" month="March"/>
</front> </front>
<seriesInfo name="Article" value="IEEE Network Magazine"/> <refcontent>IEEE Network, Volume 14, Issue 2, Pages 28-33</refcontent>
</reference> <seriesInfo name="DOI" value="10.1109/65.826369"/>
</reference>
<reference anchor="YARE95" target="http://www.cs.uccs.edu/~zbo/teaching/CS522/ <reference anchor="YARE95" target="https://ieeexplore.ieee.org/document/39
Projects/Taxonomy_Network1995.pdf"> 7042">
<front> <front>
<title>A Taxonomy for Congestion Control Algorithms in Packet Switching Ne <title>A Taxonomy for Congestion Control Algorithms in Packet
tworks</title> Switching Networks</title>
<author initials="C." surname="Yang"> <author initials="C." surname="Yang">
<organization></organization> <organization/>
</author> </author>
<author initials="A." surname="Reddy"> <author initials="A." surname="Reddy">
<organization></organization> <organization/>
</author> </author>
<date year="1995"/> <date month="August" year="1995"/>
</front> </front>
<seriesInfo name="Article" value="IEEE Network Magazine, p. 34-45"/> <refcontent>IEEE Network, Pages 34-45</refcontent>
</reference> <seriesInfo name="DOI" value="10.1109/65.397042"/>
</reference>
</references> </references>
<section anchor="CHANGES" title="Summary of Changes Since RFC 3272"> <section anchor="CHANGES" numbered="true" toc="default">
<name>Summary of Changes since RFC 3272</name>
<t>The changes to this document since <xref target="RFC3272"
format="default"/> are substantial and not easily summarized as
section-by-section changes. The material in the document has been moved
around considerably, some of it removed, and new text added.</t>
<t>The approach taken here is to list the contents of both <xref target="R
FC3272"
format="default"/>
and this document saying, respectively, where the text has
been placed and where the text came from.</t>
<section anchor="OLD" numbered="true" toc="default">
<name>RFC 3272</name>
<t>The changes to this document since RFC 3272 are substantial and not easily <ul spacing="normal">
summarized as section-by-section changes. The material in the document has <li><t>Section <xref target="RFC3272" sectionFormat="bare"
been moved around considerably, some of it removed, and new text added.</t> section="1.0">"Introduction"</xref>: Edited in place in <xref
target="INTRO" format="default"/>.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" section="1
.1">"What is Internet Traffic Engineering?"</xref>: Edited in place
in <xref target="WHATTE" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="1
.2">"Scope"</xref>: Moved to <xref target="SCOPE"
format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="1
.3">"Terminology"</xref>: Moved to <xref target="TERMS"
format="default"/> with some obsolete terms removed and a little
editing.</li>
</ul>
</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section="2.
0">"Background"</xref>: Retained as <xref target="BG"
format="default"/> with some text removed.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" section="2
.1">"Context of Internet Traffic Engineering"</xref>: Retained as
<xref target="CONTEXT" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="2
.2">"Network Context"</xref>: Rewritten as <xref target="NWCTXT"
format="default"/>.</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section
="2.3">"Problem Context"</xref>: Rewritten as <xref target="PRBCTXT"
format="default"/>.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="2.3.1">"Congestion and its Ramifications"</xref>: Retained as
<xref target="CONGEST" format="default"/>.</li>
</ul>
</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section
="2.4">"Solution Context"</xref>: Edited as <xref target="SLNCTXT"
format="default"/>.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="2.4.1">"Combating the Congestion Problem"</xref>: Reformatted as
<xref target="COMBAT" format="default"/>.</li>
</ul>
</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="2.
5">"Implementation and Operational Context"</xref>: Retained as
<xref target="IMPCTXT" format="default"/>.</li>
</ul>
</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section="3.
0">"Traffic Engineering Process Model"</xref>: Retained as <xref
target="TEPROC" format="default"/>.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" section="3.1"
>"Components of the Traffic Engineering Process Model"</xref>:
Retained as <xref target="COMPONENT"
format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="3.2"
>"Measurement"</xref>: Merged into <xref target="COMPONENT"
format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="3.3"
>"Modeling, Analysis, and Simulation"</xref>: Merged into
<xref target="COMPONENT" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="3.4"
>"Optimization"</xref>: Merged into <xref target="COMPONENT"
format="default"/>.</li>
</ul>
</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section="4.
0">"Historical Review and Recent Developments"</xref>: Retained as
<xref target="REVIEW" format="default"/>, but the very historic
aspects have been deleted.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" section="4
.1">"Traffic Engineering in Classical Telephone Networks"</xref>: Deleted.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="4
.2">"Evolution of Traffic Engineering in the Internet"</xref>: Deleted.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="4
.3">"Overlay Model"</xref>: Deleted.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="4
.4">"Constraint-Based Routing"</xref>: Retained as <xref
target="CSPF" format="default"/>, but moved into <xref
target="OTHER" format="default"/>.</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section
="4.5">"Overview of Other IETF Projects Related to Traffic
Engineering"</xref>: Retained as <xref target="OTHER" format="defa
ult"/>
with many new subsections.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="4.5.1">"Integrated Services"</xref>: Retained as <xref
target="INTSERV" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="4.5.2">"RSVP"</xref>: Retained as <xref target="RSVP"
format="default"/> with some edits.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="4.5.3">"Differentiated Services"</xref>: Retained as <xref
target="DIFFSERV" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="4.5.4">"MPLS"</xref>: Retained as <xref target="MPLS"
format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="4.5.5">"IP Performance Metrics"</xref>: Retained as <xref
target="IPPM" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="4.5.6">"Flow Measurement"</xref>: Retained as <xref target="RTFM"
format="default"/> with some reformatting.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="4.5.7">"Endpoint Congestion Management"</xref>: Retained as <xref
target="ECM" format="default"/>.</li>
</ul>
</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="4
.6">"Overview of ITU Activities Related to Traffic
Engineering"</xref>: Deleted.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="4
.7">"Content Distribution"</xref>: Retained as <xref target="CDN"
format="default"/>.</li>
</ul>
</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section="5.
0">"Taxonomy of Traffic Engineering Systems"</xref>: Retained as
<xref target="TAXI" format="default"/>.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" section="5
.1">"Time-Dependent Versus State-Dependent"</xref>: Retained as <xref
target="TIME" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="5
.2">"Offline Versus Online"</xref>: Retained as <xref target="OFFON"
format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="5
.3">"Centralized Versus Distributed"</xref>: Retained as <xref
target="CENTRAL" format="default"/> with additions.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="5
.4">"Local Versus Global"</xref>: Retained as <xref target="LOCAL"
format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="5
.5">"Prescriptive Versus Descriptive"</xref>: Retained as <xref
target="SCRIPT" format="default"/> with additions.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="5
.6">"Open-Loop Versus Closed-Loop"</xref>: Retained as <xref
target="LOOP" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="5
.7">"Tactical vs Strategic"</xref>: Retained as <xref target="TACTIC"
format="default"/>.</li>
</ul>
</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section="6.
0">"Recommendations for Internet Traffic Engineering"</xref>:
Retained as <xref target="RECO" format="default"/>.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" section="6
.1">"Generic Non-functional Recommendations"</xref>: Retained as
<xref target="HIGHOBJ" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="6
.2">"Routing Recommendations"</xref>: Retained as <xref
target="ROUTEREC" format="default"/> with edits.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="6
.3">"Traffic Mapping Recommendations"</xref>: Retained as <xref
target="MAPREC" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="6
.4">"Measurement Recommendations"</xref>: Retained as <xref
target="MSRREC" format="default"/>.</li>
<li><t>Section <xref target="RFC3272" sectionFormat="bare" section
="6.5">"Network Survivability"</xref>: Retained as <xref
target="SURVIVE" format="default"/>.</t>
<ul spacing="normal">
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="6.5.1">"Survivability in MPLS Based Networks"</xref>: Retained as
<xref target="SRVMPLS" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" sectio
n="6.5.2">"Protection Option"</xref>: Retained as <xref
target="PROTECT" format="default"/>.</li>
</ul>
</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="6
.6">"Traffic Engineering in Diffserv Environments"</xref>: Retained
as <xref target="TEDIFFSRV" format="default"/> with edits.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="6
.7">"Network Controllability"</xref>: Retained as <xref
target="CONTROL" format="default"/>.</li>
</ul>
</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="7.0">
"Inter-Domain Considerations"</xref>: Retained as <xref
target="INTER" format="default"/>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="8.0">"
Overview of Contemporary TE Practices in Operational IP
Networks"</xref>: Retained as <xref target="PRACTICE" format="default"/
>.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="9.0">
"Conclusion"</xref>: Removed.</li>
<li>Section <xref target="RFC3272" sectionFormat="bare" section="10.0">
"Security Considerations"</xref>: Retained as <xref target="SECURE"
format="default"/> with considerable new text.</li>
</ul>
<t>The approach taken here is to list the table of content of both the previou </section>
s <section anchor="NEW" numbered="true" toc="default">
RFC and this document saying, respectively, where the text has been placed <name>This Document</name>
and where the text came from.</t> <ul spacing="normal">
<li>
<t><xref target="INTRO" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="1"/>. </t>
<ul spacing="normal">
<li>
<xref target="WHATTE" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="1.1"/>.</li>
<li>
<xref target="COMPONENTS" format="default"/>: New for this docum
ent.</li>
<li>
<xref target="SCOPE" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="1.2"/>.</li>
<li>
<xref target="TERMS" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="1.3"/>.</li>
</ul>
</li>
<li>
<t><xref target="BG" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="2"/>.
</t>
<ul spacing="normal">
<li>
<xref target="CONTEXT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="2.1"/>.</li>
<li>
<xref target="NWCTXT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="2.2"/>.</li>
<li>
<t><xref target="PRBCTXT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="2.3"/>.
</t>
<ul spacing="normal">
<li>
<xref target="CONGEST" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="2.3.1"/>.</li>
</ul>
</li>
<li>
<t><xref target="SLNCTXT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="2.4"/>.</t>
<ul spacing="normal">
<li>
<xref target="COMBAT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="2.4.1"/>.</li>
</ul>
</li>
<li>
<xref target="IMPCTXT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="2.5"/>.</li>
</ul>
</li>
<li>
<t><xref target="TEPROC" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="3"/>.
</t>
<ul spacing="normal">
<li><xref target="COMPONENT" format="default"/>: Based on
Sections <xref target="RFC3272" sectionFormat="bare"
section="3.1"/>, <xref target="RFC3272" sectionFormat="bare"
section="3.2"/>, <xref target="RFC3272" sectionFormat="bare"
section="3.3"/>, and <xref target="RFC3272" sectionFormat="bare"
section="3.4"/> of <xref target="RFC3272"
format="default"/>.</li>
</ul>
</li>
<li>
<t><xref target="TAXI" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5"/>.
</t>
<ul spacing="normal">
<li>
<xref target="TIME" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5.1"/>.</li>
<li>
<xref target="OFFON" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5.2"/>.</li>
<li>
<t><xref target="CENTRAL" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5.3"/>.
</t>
<ul spacing="normal">
<li>
<xref target="HYBRID" format="default"/>: New for this docum
ent.</li>
<li>
<xref target="SDN" format="default"/>: New for this document
.</li>
</ul>
</li>
<li>
<xref target="LOCAL" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5.4"/>.</li>
<li>
<t><xref target="SCRIPT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5.5"/>.
</t>
<ul spacing="normal">
<li>
<xref target="INTENT" format="default"/>: New for this docum
ent.</li>
</ul>
</li>
<li>
<xref target="LOOP" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5.6"/>.</li>
<li>
<xref target="TACTIC" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="5.7"/>.</li>
</ul>
</li>
<li>
<t><xref target="REVIEW" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="4"/>.
</t>
<ul spacing="normal">
<li>
<t><xref target="OTHER" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="4.5"/>.
</t>
<ul spacing="normal">
<li>
<xref target="INTSERV" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="4.5.1"/>.</li>
<li>
<xref target="DIFFSERV" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="4.5.3"/>.</li>
<li>
<xref target="SRPolicy" format="default"/>: New for this doc
ument.</li>
<li>
<xref target="QUIC" format="default"/>: New for this documen
t.</li>
<li>
<xref target="DETNET" format="default"/>: New for this docum
ent.</li>
<li>
<xref target="ALTO" format="default"/>: New for this documen
t.</li>
<li>
<xref target="ACTN" format="default"/>: New for this documen
t.</li>
<li>
<xref target="SLICE" format="default"/>: New for this docume
nt.</li>
<li>
<t><xref target="CSPF" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="4.4"/>.
</t>
<ul spacing="normal">
<li>
<xref target="FLEX" format="default"/>: New for this doc
ument.</li>
</ul>
</li>
<li>
<xref target="RSVP" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="4.5.2"/>.</li>
<li>
<xref target="MPLS" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="4.5.4"/>.</li>
<li>
<xref target="RSVP-TE" format="default"/>: New for this docu
ment.</li>
<li>
<xref target="GMPLS" format="default"/>: New for this docume
nt.</li>
<li>
<xref target="IPPM" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="4.5.5"/>.</li>
<li>
<xref target="RTFM" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="4.5.6"/>.</li>
<li>
<xref target="ECM" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="4.5.7"/>.</li>
<li>
<xref target="IGPTE" format="default"/>: New for this docume
nt.</li>
<li>
<xref target="BGPLS" format="default"/>: New for this docume
nt.</li>
<li>
<xref target="PCE" format="default"/>: New for this document
.</li>
<li>
<xref target="SR" format="default"/>: New for this document.
</li>
<li>
<xref target="BIER-TE" format="default"/>: New for this docu
ment.</li>
<li>
<xref target="STATE" format="default"/>: New for this docume
nt.</li>
<li>
<xref target="SYSMAN" format="default"/>: New for this docum
ent.</li>
</ul>
</li>
<li>
<xref target="CDN" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="4.7"/>.</li>
</ul>
</li>
<li>
<t><xref target="RECO" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6"/>.
</t>
<ul spacing="normal">
<li>
<xref target="HIGHOBJ" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6.1"/>.</li>
<li>
<xref target="ROUTEREC" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6.2"/>.</li>
<li>
<xref target="MAPREC" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6.3"/>.</li>
<li>
<xref target="MSRREC" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6.4"/>.</li>
<li>
<xref target="POLICE" format="default"/>: New for this document.
</li>
<li>
<t><xref target="SURVIVE" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6.5"/>.
</t>
<ul spacing="normal">
<li>
<xref target="SRVMPLS" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="6.5.1"/>.</li>
<li>
<xref target="PROTECT" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of"
section="6.5.2"/>.</li>
</ul>
</li>
<li>
<xref target="ML" format="default"/>: New for this document.</li
>
<li>
<xref target="TEDIFFSRV" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6.6"/>.</li>
<li>
<xref target="CONTROL" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="6.7"/>.</li>
</ul>
</li>
<li>
<xref target="INTER" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="7"/>.</li>
<li>
<xref target="PRACTICE" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="8"/>.</li>
<li>
<xref target="SECURE" format="default"/>: Based on <xref
target="RFC3272" sectionFormat="of" section="10"/>.</li>
</ul>
</section>
</section>
<section anchor="OLD" title="RFC 3272"> <section anchor="ACKN" numbered="false" toc="default">
<t><list style="hanging"> <name>Acknowledgments</name>
<t hangText="1.0 Introduction:">Edited in place in <xref target="INTRO"
/>.
<list style="hanging">
<t hangText="1.1 What is Internet Traffic Engineering?:">Edited in
place in <xref target="WHATTE" />.</t>
<t hangText="1.2 Scope:">Moved to <xref target="SCOPE" />.</t>
<t hangText="1.3 Terminology:">Moved to <xref target="TERMS" /> wi
th some obsolete terms removed
and a little editing.</t>
</list></t>
<t hangText="2.0 Background:">Retained as <xref target="BG" /> with some <t>Much of the text in this document is derived from <xref
text removed. target="RFC3272" format="default"/>. The editor and contributors to
<list style="hanging"> this document would like to express their gratitude to all involved in
<t hangText="2.1 Context of Internet Traffic Engineering:">Retaine that work. Although the source text has been edited in the production
d as <xref target="CONTEXT" />.</t> of this document, the original authors should be considered as
<t hangText="2.2 Network Context:">Rewritten as <xref target="NWCT contributors to this work. They were:</t>
XT" />.</t>
<t hangText="2.3 Problem Context:">Rewritten as <xref target="PRBC
TXT" />.
<list style="hanging">
<t hangText="2.3.1 Congestion and its Ramifications:">Retain
ed as <xref target="CONGEST" />.</t>
</list></t>
<t hangText="2.4 Solution Context:">Edited as <xref target="SLNCTX
T" />.
<list style="hanging">
<t hangText="2.4.1 Combating the Congestion Problem:">Reform
atted as <xref target="COMBAT" />.</t>
</list></t>
<t hangText="2.5 Implementation and Operational Context:">Retained
as <xref target="IMPCTXT" />.</t>
</list></t>
<t hangText="3.0 Traffic Engineering Process Model:">Retained as <xref t <contact fullname="Daniel O. Awduche">
arget="TEPROC" />. <organization>Movaz Networks</organization>
<list style="hanging"> </contact>
<t hangText="3.1 Components of the Traffic Engineering Process Mod
el:">Retained as <xref target="COMPONENT" />.</t>
<t hangText="3.2 Measurement:">Merged into <xref target="COMPONENT
" />.</t>
<t hangText="3.3 Modeling, Analysis, and Simulation:">Merged into
<xref target="COMPONENT" />.</t>
<t hangText="3.4 Optimization:">Merged into <xref target="COMPONEN
T" />.</t>
</list></t>
<t hangText="4.0 Historical Review and Recent Developments:">Retained as <contact fullname="Angela Chiu">
<xref target="REVIEW" />, but the very <organization>Celion Networks</organization>
historic aspects have been deleted. </contact>
<list style="hanging">
<t hangText="4.1 Traffic Engineering in Classical Telephone Networ
ks:">Deleted.</t>
<t hangText="4.2 Evolution of Traffic Engineering in the Internet:
">Deleted.</t>
<t hangText="4.3 Overlay Model:">Deleted.</t>
<t hangText="4.4 Constraint-Based Routing:">Retained as <xref targ
et="CSPF" />, but moved into <xref target="OTHER" />.</t>
<t hangText="4.5 Overview of Other IETF Projects Related to Traffi
c Engineering:">Retained as <xref target="OTHER" />
with many new subsections.
<list style="hanging">
<t hangText="4.5.1 Integrated Services:">Retained as <xref t
arget="INTSERV" />.</t>
<t hangText="4.5.2 RSVP:">Retained as <xref target="RSVP" />
with some edits.</t>
<t hangText="4.5.3 Differentiated Services:">Retained as <xr
ef target="DIFFSERV" />.</t>
<t hangText="4.5.4 MPLS:">Retained as <xref target="MPLS" />
.</t>
<t hangText="4.5.5 IP Performance Metrics:">Retained as <xre
f target="IPPM" />.</t>
<t hangText="4.5.6 Flow Measurement:">Retained as <xref targ
et="RTFM" /> with some reformatting.</t>
<t hangText="4.5.7 Endpoint Congestion Management:">Retained
as <xref target="ECM" />.</t>
</list></t>
<t hangText="4.6 Overview of ITU Activities Related to Traffic Eng
ineering:">Deleted.</t>
<t hangText="4.7 Content Distribution:">Retained as <xref target="
CDN" />.</t>
</list></t>
<t hangText="5.0 Taxonomy of Traffic Engineering Systems:">Retained as < <contact fullname="Anwar Elwalid">
xref target="TAXI" />. <organization>Lucent Technologies</organization>
<list style="hanging"> </contact>
<t hangText="5.1 Time-Dependent Versus State-Dependent:">Retained
as <xref target="TIME" />.</t>
<t hangText="5.2 Offline Versus Online:">Retained as <xref target=
"OFFON" />.</t>
<t hangText="5.3 Centralized Versus Distributed:">Retained as <xre
f target="CENTRAL" /> with additions.</t>
<t hangText="5.4 Local Versus Global:">Retained as <xref target="L
OCAL" />.</t>
<t hangText="5.5 Prescriptive Versus Descriptive:">Retained as <xr
ef target="SCRIPT" /> with additions.</t>
<t hangText="5.6 Open-Loop Versus Closed-Loop:">Retained as <xref
target="LOOP" />.</t>
<t hangText="5.7 Tactical vs Strategic:">Retained as <xref target=
"TACTIC" />.</t>
</list></t>
<t hangText="6.0 Recommendations for Internet Traffic Engineering:">Reta <contact fullname="Indra Widjaja">
ined as <xref target="RECO" />. <organization>Bell Labs, Lucent Technologies</organization>
<list style="hanging"> </contact>
<t hangText="6.1 Generic Non-functional Recommendations:">Retained
as <xref target="HIGHOBJ" />.</t>
<t hangText="6.2 Routing Recommendations:">Retained as <xref targe
t="ROUTEREC" /> with edits.</t>
<t hangText="6.3 Traffic Mapping Recommendations:">Retained as <xr
ef target="MAPREC" />.</t>
<t hangText="6.4 Measurement Recommendations:">Retained as <xref t
arget="MSRREC" />.</t>
<t hangText="6.5 Network Survivability:">Retained as <xref target=
"SURVIVE" />.
<list style="hanging">
<t hangText="6.5.1 Survivability in MPLS Based Networks:">Re
tained as <xref target="SRVMPLS" />.</t>
<t hangText="6.5.2 Protection Option:">Retained as <xref tar
get="PROTECT" />.</t>
</list></t>
<t hangText="6.6 Traffic Engineering in Diffserv Environments:">Re
tained as <xref target="TEDIFFSRV" /> with edits.</t>
<t hangText="6.7 Network Controllability:">Retained as <xref targe
t="CONTROL" />.</t>
</list></t>
<t hangText="7.0 Inter-Domain Considerations:">Retained as <xref target= <contact fullname="XiPeng Xiao">
"INTER" />.</t> <organization>Redback Networks</organization>
<t hangText="8.0 Overview of Contemporary TE Practices in Operational IP </contact>
Networks:">Retained as <xref target="PRACTICE" />.</t>
<t hangText="9.0 Conclusion:">Removed.</t>
<t hangText="10.0 Security Considerations:">Retained as <xref target="SE
CURE" /> with considerable new text.</t>
</list></t>
</section> <t>The acknowledgements in <xref target="RFC3272" format="default"/>
were as below. All people who helped in the production of that document
also need to be thanked for the carry-over into this new document.</t>
<section anchor="NEW" title="This Document"> <blockquote><t>The authors would like to thank <contact fullname="Jim
Boyle"/> for inputs on the recommendations section, <contact
fullname="Francois Le Faucheur"/> for inputs on Diffserv aspects,
<contact fullname="Blaine Christian"/> for inputs on measurement,
<contact fullname="Gerald Ash"/> for inputs on routing in telephone
networks and for text on event-dependent TE methods, <contact
fullname="Steven Wright"/> for inputs on network controllability, and
<contact fullname="Jonathan Aufderheide"/> for inputs on inter-domain TE
with BGP. Special thanks to <contact fullname="Randy Bush"/> for
proposing the TE taxonomy based on "tactical vs strategic" methods. The
subsection describing an "Overview of ITU Activities Related to Traffic
Engineering" was adapted from a contribution by <contact
fullname="Waisum Lai"/>. Useful feedback and pointers to relevant
materials were provided by <contact fullname="J. Noel Chiappa"/>.
Additional comments were provided by <contact fullname="Glenn
Grotefeld"/> during the working last call process. Finally, the authors
would like to thank <contact fullname="Ed Kern"/>, the TEWG co-chair,
for his comments and support.</t></blockquote>
<t><list style="symbols"> <t>The early draft versions of this document were produced by the TEAS Wor
<t><xref target="INTRO" />: Based on Section 1 of RFC 3272. king
<list style="symbols"> Group's RFC3272bis Design Team. The full list of members of this team
<t><xref target="WHATTE" />: Based on Section 1.1 of RFC 3272.</t> is:</t>
<t><xref target="COMPONENTS" />: New for this document.</t> <ul empty="true" spacing="compact" bare="false" indent="3">
<t><xref target="SCOPE" />: Based on Section 1.2 of RFC 3272.</t> <li><t><contact fullname="Acee Lindem"/></t></li>
<t><xref target="TERMS" />: Based on Section 1.3 of RFC 3272.</t> <li><t><contact fullname="Adrian Farrel"/></t></li>
</list></t> <li><t><contact fullname="Aijun Wang"/></t></li>
<li><t><contact fullname="Daniele Ceccarelli"/></t></li>
<li><t><contact fullname="Dieter Beller"/></t></li>
<li><t><contact fullname="Jeff Tantsura"/></t></li>
<li><t><contact fullname="Julien Meuric"/></t></li>
<li><t><contact fullname="Liu Hua"/></t></li>
<li><t><contact fullname="Loa Andersson"/></t></li>
<li><t><contact fullname="Luis Miguel Contreras"/></t></li>
<li><t><contact fullname="Martin Horneffer"/></t></li>
<li><t><contact fullname="Tarek Saad"/></t></li>
<li><t><contact fullname="Xufeng Liu"/></t></li>
</ul>
<t>The production of this document includes a fix to the original text
resulting from an errata report #309 <xref target="Err309"/> by <contact f
ullname="Jean-Michel
Grimaldi"/>.</t>
<t>The editor of this document would also like to thank <contact
fullname="Dhruv Dhody"/>, <contact fullname="Gyan Mishra"/>, <contact
fullname="Joel Halpern"/>, <contact fullname="Dave Taht"/>, <contact
fullname="John Scudder"/>, <contact fullname="Rich Salz"/>, <contact
fullname="Behcet Sarikaya"/>, <contact fullname="Bob Briscoe"/>,
<contact fullname="Erik Kline"/>, <contact fullname="Jim Guichard"/>,
<contact fullname="Martin Duke"/>, and <contact fullname="Roman
Danyliw"/> for review comments.</t>
<t>This work is partially supported by the European Commission under
Horizon 2020 grant agreement number 101015857 Secured autonomic traffic
management for a Tera of SDN flows (Teraflow).</t>
</section>
<t><xref target="BG" />: Based on Section 2. of RFC 3272. <section anchor="CONTRIB" numbered="false" toc="default">
<list style="symbols"> <name>Contributors</name>
<t><xref target="CONTEXT" />: Based on Section 2.1 of RFC 3272.</t <t>The following people contributed substantive text to this
> document:</t>
<t><xref target="NWCTXT" />: Based on Section 2.2 of RFC 3272.</t>
<t><xref target="PRBCTXT" />: Based on Section 2.3 of RFC 3272.
<list style="symbols">
<t><xref target="CONGEST" />: Based on Section 2.3.1 of RFC 327
2.</t>
</list></t>
<t><xref target="SLNCTXT" />: Based on Section 2.4 of RFC 3272.
<list style="symbols">
<t><xref target="COMBAT" />: Based on Section 2.4.1 of RFC 327<
/t>
</list></t>
<t><xref target="IMPCTXT" />: Based on Section 2.5 of RFC 3272.</t
>
</list></t>
<t><xref target="TEPROC" />: Based on Section 3 of RFC 3272. <contact fullname="Gert Grammel">
<list style="symbols"> <address>
<t><xref target="COMPONENT" />: Based on Sections 3.1, 3.2, 3.3, a <email>ggrammel@juniper.net</email>
nd 3.4 of RFC 3272.</t> </address>
</list></t> </contact>
<t><xref target="TAXI" />: Based on Section 5 of RFC 3272. <contact fullname="Loa Andersson">
<list style="symbols"> <address>
<t><xref target="TIME" />: Based on Section 5.1 of RFC 3272.</t> <email>loa@pi.nu</email>
<t><xref target="OFFON" />: Based on Section 5.2 of RFC 3272.</t> </address>
<t><xref target="CENTRAL" />: Based on Section 5.3 of RFC 3272. </contact>
<list style="symbols">
<t><xref target="HYBRID" />: New for this document.</t>
<t><xref target="SDN" />: New for this document.</t>
</list></t>
<t><xref target="LOCAL" />: Based on Section 5.4 of RFC 3272.</t>
<t><xref target="SCRIPT" />: Based on Section 5.5 of RFC 3272.
<list style="symbols">
<t><xref target="INTENT" />: New for this document.</t>
</list></t>
<t><xref target="LOOP" />: Based on Section 5.6 of RFC 3272.</t>
<t><xref target="TACTIC" />: Based on Section 5.7 of RFC 3272.</t>
</list></t>
<t><xref target="REVIEW" />: Based on Section 4 of RFC 3272. <contact fullname="Xufeng Liu">
<list style="symbols"> <address>
<t><xref target="OTHER" />: Based on Section 4.5 of RFC 3272. <email>xufeng.liu.ietf@gmail.com</email>
<list style="symbols"> </address>
<t><xref target="INTSERV" />: Based on Section 4.5.1 of RFC 327 </contact>
2.</t>
<t><xref target="DIFFSERV" />: Based on Section 4.5.3 of RFC 32
72.</t>
<t><xref target="SRPolicy" />: New for this document.</t>
<t><xref target="QUIC" />: New for this document.</t>
<t><xref target="DETNET" />: New for this document.</t>
<t><xref target="ALTO" />: New for this document.</t>
<t><xref target="ACTN" />: New for this document.</t>
<t><xref target="SLICE" />: New for this document.</t>
<t><xref target="CSPF" />: Based on Section 4.4 of RFC 3272.
<list style="symbols">
<t><xref target="FLEX" />: New for this document.</t>
</list></t>
<t><xref target="RSVP" />: Based on Section 4.5.2 of RFC 3272.<
/t>
<t><xref target="MPLS" />: Based on Section 4.5.4 of RFC 3272.<
/t>
<t><xref target="RSVP-TE" />: New for this document.</t>
<t><xref target="GMPLS" />: New for this document.</t>
<t><xref target="IPPM" />: Based on Section 4.5.5 of RFC 3272.<
/t>
<t><xref target="RTFM" />: Based on Section 4.5.6 of RFC 3272.<
/t>
<t><xref target="ECM" />: Based on Section 4.5.7 of RFC 3272.</
t>
<t><xref target="IGPTE" />: New for this document.</t>
<t><xref target="BGPLS" />: New for this document.</t>
<t><xref target="PCE" />: New for this document.</t>
<t><xref target="SR" />: New for this document.</t>
<t><xref target="BIER-TE" />: New for this document.</t>
<t><xref target="STATE" />: New for this document.</t>
<t><xref target="SYSMAN" />: New for this document.</t>
</list></t>
<t><xref target="CDN" />: Based on Section 4.7 of RFC 3272.</t>
</list></t>
<t><xref target="RECO" />: Based on Section 6 of RFC 3272. <contact fullname="Lou Berger">
<list style="symbols"> <address>
<t><xref target="HIGHOBJ" />: Based on Section 6.1 of RFC 3272.</t <email>lberger@labn.net</email>
> </address>
<t><xref target="ROUTEREC" />: Based on Section 6.2 of RFC 3272.</ </contact>
t>
<t><xref target="MAPREC" />: Based on Section 6.3 of RFC 3272.</t>
<t><xref target="MSRREC" />: Based on Section 6.4 of RFC 3272.</t>
<t><xref target="POLICE" />: New for this document.</t>
<t><xref target="SURVIVE" />: Based on Section 6.5 of RFC 3272.
<list style="symbols">
<t><xref target="SRVMPLS" />: Based on Section 6.5.1 of RFC 327
2.</t>
<t><xref target="PROTECT" />: Based on Section 6.5.2 of RFC 327
2.</t>
</list></t>
<t><xref target="ML" />: New for this document.</t>
<t><xref target="TEDIFFSRV" />: Based on Section 6.6. of RFC 3272.
</t>
<t><xref target="CONTROL" />: Based on Section 6.7 of RFC 3272.</t
>
</list></t>
<t><xref target="INTER" />: Based on Section 7 of RFC 3272.</t> <contact fullname="Jeff Tantsura">
<address>
<email>jefftant.ietf@gmail.com</email>
</address>
</contact>
<t><xref target="PRACTICE" />: Based on Section 8 of RFC 3272.</t> <contact fullname="Daniel King">
<address>
<email>daniel@olddog.co.uk</email>
</address>
</contact>
<t><xref target="SECURE" />: Based on Section 10 of RFC 3272.</t> <contact fullname="Boris Hassanov">
<address>
<email>bhassanov@yandex-team.ru</email>
</address>
</contact>
</list></t> <contact fullname="Kiran Makhijani">
<address>
<email>kiranm@futurewei.com</email>
</address>
</contact>
</section> <contact fullname="Dhruv Dhody">
<address>
<email>dhruv.ietf@gmail.com</email>
</address>
</contact>
</section> <contact fullname="Mohamed Boucadair">
<address>
<email>mohamed.boucadair@orange.com</email>
</address>
</contact>
</back> </section>
</back>
</rfc> </rfc>
 End of changes. 145 change blocks. 
4872 lines changed or deleted 4414 lines changed or added

This html diff was produced by rfcdiff 1.48.