rfc9332.original.xml   rfc9332.xml 
<?xml version='1.0' encoding='utf-8'?> <?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE rfc [ <!DOCTYPE rfc [
<!ENTITY nbsp "&#160;"> <!ENTITY nbsp "&#160;">
<!ENTITY zwsp "&#8203;"> <!ENTITY zwsp "&#8203;">
<!ENTITY nbhy "&#8209;"> <!ENTITY nbhy "&#8209;">
<!ENTITY wj "&#8288;"> <!ENTITY wj "&#8288;">
]> ]>
<!-- This template is for creating an Internet Draft using xml2rfc,
which is available here: http://xml.resource.org. -->
<?xml-stylesheet type='text/xsl' href='rfc2629.xslt' ?>
<!-- used by XSLT processors -->
<!-- For a complete list and description of processing instructions (PIs),
please see http://xml.resource.org/authoring/README.html. -->
<!-- Below are generally applicable Processing Instructions (PIs) that most I-Ds
might want to use.
(Here they are set differently than their defaults in xml2rfc v1.32) -->
<?rfc strict="yes" ?>
<!-- give errors regarding ID-nits and DTD validation -->
<!-- control the table of contents (ToC) -->
<?rfc toc="yes"?>
<!-- generate a ToC -->
<?rfc tocdepth="4"?>
<!-- the number of levels of subsections in ToC. default: 3 -->
<!-- control references -->
<?rfc symrefs="yes"?>
<!-- use symbolic references tags, i.e, [RFC2119] instead of [1] -->
<?rfc sortrefs="yes" ?>
<!-- sort the reference entries alphabetically -->
<!-- control vertical white space
(using these PIs as follows is recommended by the RFC Editor) -->
<?rfc compact="yes" ?>
<!-- do not start each main section on a new page -->
<?rfc subcompact="no" ?>
<!-- keep one blank line between list items -->
<!-- end of list of popular I-D processing instructions -->
<rfc xmlns:xi="http://www.w3.org/2001/XInclude" category="exp" docName="draft-ie
tf-tsvwg-aqm-dualq-coupled-25" ipr="trust200902" updates="" obsoletes="" submiss
ionType="IETF" xml:lang="en" tocInclude="true" tocDepth="4" symRefs="true" sortR
efs="true" version="3">
<!-- xml2rfc v2v3 conversion 3.14.1 -->
<!-- category values: std, bcp, info, exp, and historic
ipr values: trust200902, noModificationTrust200902, noDerivativesTrust200902
,
or pre5378Trust200902
you can add the attributes updates="NNNN" and obsoletes="NNNN"
they will automatically be output with "(if approved)" -->
<!-- ***** FRONT MATTER ***** --> <rfc xmlns:xi="http://www.w3.org/2001/XInclude" submissionType="IETF" category="
exp" consensus="true" docName="draft-ietf-tsvwg-aqm-dualq-coupled-25" number="93
32" ipr="trust200902" updates="" obsoletes="" xml:lang="en" tocInclude="true" to
cDepth="4"
symRefs="true" sortRefs="true" version="3">
<!-- xml2rfc v2v3 conversion 3.14.1 -->
<front> <front>
<!-- The abbreviated title is used in the page header - it is only necessary <title abbrev="DualQ Coupled AQMs">Dual-Queue Coupled Active Queue Managemen
if the t (AQM) for Low Latency, Low Loss, and Scalable Throughput (L4S)</title>
full title is longer than 39 characters --> <seriesInfo name="RFC" value="9332"/>
<title abbrev="DualQ Coupled AQMs">DualQ Coupled AQMs for Low Latency, Low
Loss and Scalable Throughput (L4S)</title>
<seriesInfo name="Internet-Draft" value="draft-ietf-tsvwg-aqm-dualq-coupled-
25"/>
<author fullname="Koen De Schepper" initials="K." surname="De Schepper"> <author fullname="Koen De Schepper" initials="K." surname="De Schepper">
<organization>Nokia Bell Labs</organization> <organization>Nokia Bell Labs</organization>
<address> <address>
<postal> <postal>
<street/> <city>Antwerp</city>
<city>Antwerp</city>
<country>Belgium</country> <country>Belgium</country>
</postal> </postal>
<email>koen.de_schepper@nokia.com</email> <email>koen.de_schepper@nokia.com</email>
<uri>https://www.bell-labs.com/about/researcher-profiles/koende_schepper /</uri> <uri>https://www.bell-labs.com/about/researcher-profiles/koende_schepper /</uri>
</address> </address>
</author> </author>
<author fullname="Bob Briscoe" initials="B." role="editor" surname="Briscoe" > <author fullname="Bob Briscoe" initials="B." role="editor" surname="Briscoe" >
<organization>Independent</organization> <organization>Independent</organization>
<address> <address>
<postal> <postal>
<street/> <country>United Kingdom</country>
<country>UK</country>
</postal> </postal>
<email>ietf@bobbriscoe.net</email> <email>ietf@bobbriscoe.net</email>
<uri>https://bobbriscoe.net/</uri> <uri>https://bobbriscoe.net/</uri>
</address> </address>
</author> </author>
<author fullname="Greg White" initials="G." surname="White"> <author fullname="Greg White" initials="G." surname="White">
<organization>CableLabs</organization> <organization>CableLabs</organization>
<address> <address>
<postal> <postal>
<street/> <city>Louisville</city>
<city>Louisville, CO</city> <region>CO</region>
<country>US</country> <country>United States of America</country>
</postal> </postal>
<email>G.White@CableLabs.com</email> <email>G.White@CableLabs.com</email>
</address> </address>
</author> </author>
<!-- <author fullname="Olga Albisser" initials="O." surname="Albisser"> <date year="2023" month="January" />
<organization>Simula Research Lab</organization>
<address>
<postal>
<street/>
<city>Lysaker</city>
<country>Norway</country>
</postal>
<email>olga@albisser.org</email>
<uri>https://www.simula.no/people/olgabo</uri>
</address>
</author>
<author fullname="Ing Jyh Tsang" initials="I." surname="Tsang">
<organization>Nokia</organization>
<address>
<postal>
<street/>
<city>Antwerp</city>
<country>Belgium</country> <area>tsv</area>
</postal> <workgroup>tsvwg</workgroup>
<email>ing-jyh.tsang@nokia.com</email> <keyword>Performance</keyword>
</address> <keyword>Queuing Delay</keyword>
</author> <keyword>One Way Delay</keyword>
<keyword>Round-Trip Time</keyword>
<keyword>RTT</keyword>
<keyword>Jitter</keyword>
<keyword>Congestion Control</keyword>
<keyword>Congestion Avoidance</keyword>
<keyword>Quality of Service</keyword>
<keyword>QoS</keyword>
<keyword>Quality of Experience</keyword>
<keyword>QoE</keyword>
<keyword>Active Queue Management</keyword>
<keyword>AQM</keyword>
<keyword>Explicit Congestion Notification</keyword>
<keyword>ECN</keyword>
<keyword>Pacing</keyword>
<keyword>Burstiness</keyword>
<date month="" year=""/>
<area>Transport</area>
<workgroup>Transport Area working group (tsvwg)</workgroup>
<keyword>Internet-Draft</keyword>
<keyword>I-D</keyword>
<abstract> <abstract>
<t>This specification defines a framework for coupling the Active Queue <t>This specification defines a framework for coupling the Active Queue
Management (AQM) algorithms in two queues intended for flows with Management (AQM) algorithms in two queues intended for flows with
different responses to congestion. This provides a way for the Internet different responses to congestion. This provides a way for the Internet
to transition from the scaling problems of standard TCP Reno-friendly to transition from the scaling problems of standard TCP-Reno-friendly
('Classic') congestion controls to the family of 'Scalable' congestion ('Classic') congestion controls to the family of 'Scalable' congestion
controls. These are designed for consistently very Low queuing Latency, controls. These are designed for consistently very low queuing latency,
very Low congestion Loss and Scaling of per-flow throughput (L4S) by very low congestion loss, and scaling of per-flow throughput by
using Explicit Congestion Notification (ECN) in a modified way. Until using Explicit Congestion Notification (ECN) in a modified way. Until
the Coupled DualQ, these scalable L4S congestion controls could only be the Coupled Dual Queue (DualQ), these Scalable L4S congestion controls cou ld only be
deployed where a clean-slate environment could be arranged, such as in deployed where a clean-slate environment could be arranged, such as in
private data centres.</t> private data centres.</t>
<t>The specification first explains how a Coupled DualQ works. It then
<t>This specification first explains how a Coupled DualQ works. It then
gives the normative requirements that are necessary for it to work well. gives the normative requirements that are necessary for it to work well.
All this is independent of which two AQMs are used, but pseudocode All this is independent of which two AQMs are used, but pseudocode
examples of specific AQMs are given in appendices.</t> examples of specific AQMs are given in appendices.</t>
</abstract> </abstract>
</front> </front>
<middle> <middle>
<section anchor="dualq_intro" numbered="true" toc="default"> <section anchor="dualq_intro" numbered="true" toc="default">
<name>Introduction</name> <name>Introduction</name>
<t>This document specifies a framework for DualQ Coupled AQMs, which can <t>This document specifies a framework for DualQ Coupled AQMs, which can
serve as the network part of the L4S architecture <xref target="I-D.ietf-t serve as the network part of the L4S architecture <xref target="RFC9330" f
svwg-l4s-arch" format="default"/>. A Coupled DualQ AQM consists of two ormat="default"/>. A DualQ Coupled AQM consists of two
queues; L4S and Classic. The L4S queue is intended for Scalable queues: L4S and Classic. The L4S queue is intended for Scalable
congestion controls that can maintain very low queuing latency congestion controls that can maintain very low queuing latency
(sub-millisecond on average) and high throughput at the same time. The (sub-millisecond on average) and high throughput at the same time. The
Coupled DualQ acts like a semi-permeable membrane: the L4S queue Coupled DualQ acts like a semi-permeable membrane: the L4S queue
isolates the sub-millisecond average queuing delay of L4S from Classic isolates the sub-millisecond average queuing delay of L4S from Classic
latency; while the coupling between the queues pools the capacity latency, while the coupling between the queues pools the capacity
between both queues so that ad hoc numbers of capacity-seeking between both queues so that ad hoc numbers of capacity-seeking
applications all sharing the same capacity can have roughly equivalent applications all sharing the same capacity can have roughly equivalent
throughput per flow, whichever queue they use. The DualQ achieves this throughput per flow, whichever queue they use. The DualQ achieves this
indirectly, without having to inspect transport layer flow identifiers indirectly, without having to inspect transport-layer flow identifiers
and without compromising the performance of the Classic traffic, and without compromising the performance of the Classic traffic,
relative to a single queue. The DualQ design has low complexity and relative to a single queue. The DualQ design has low complexity and
requires no configuration for the public Internet.</t> requires no configuration for the public Internet.</t>
<section anchor="dualq_problem" numbered="true" toc="default"> <section anchor="dualq_problem" numbered="true" toc="default">
<name>Outline of the Problem</name> <name>Outline of the Problem</name>
<t>Latency is becoming the critical performance factor for many <t>Latency is becoming the critical performance factor for many
(most?) applications on the public Internet, e.g. interactive (perhaps most) applications on the public Internet, e.g., interactive
Web, Web services, voice, conversational video, interactive video, web, web services, voice, conversational video, interactive video,
interactive remote presence, instant messaging, online gaming, remote interactive remote presence, instant messaging, online gaming, remote
desktop, cloud-based applications, and video-assisted remote control desktop, cloud-based applications, and video-assisted remote control
of machinery and industrial processes. Once access network bit rates of machinery and industrial processes. Once access network bitrates
reach levels now common in the developed world, further increases reach levels now common in the developed world, further increases
offer diminishing returns unless latency is also addressed <xref target= "Dukkipati06" format="default"/>. In the last decade or so, much has been done offer diminishing returns unless latency is also addressed <xref target= "Dukkipati06" format="default"/>. In the last decade or so, much has been done
to reduce propagation time by placing caches or servers closer to to reduce propagation time by placing caches or servers closer to
users. However, queuing remains a major intermittent component of users. However, queuing remains a major intermittent component of
latency.</t> latency.</t>
<t>Traditionally very low latency has only been available for a few <t>Previously, very low latency has only been available for a few
selected low rate applications, that confine their sending rate within selected low-rate applications, that confine their sending rate within
a specially carved-off portion of capacity, which is prioritized over a specially carved-off portion of capacity, which is prioritized over
other traffic, e.g. Diffserv EF <xref target="RFC3246" format="default"/ other traffic, e.g., Diffserv Expedited Forwarding (EF) <xref target="RF
>. Up C3246" format="default"/>. Up
to now it has not been possible to allow any number of low latency, to now, it has not been possible to allow any number of low-latency,
high throughput applications to seek to fully utilize available high throughput applications to seek to fully utilize available
capacity, because the capacity-seeking process itself causes too much capacity, because the capacity-seeking process itself causes too much
queuing delay.</t> queuing delay.</t>
<t>To reduce this queuing delay caused by the capacity seeking
process, changes either to the network alone or to end-systems alone <t>To reduce this queuing delay caused by the capacity-seeking
process, changes either to the network alone or to end systems alone
are in progress. L4S involves a recognition that both approaches are are in progress. L4S involves a recognition that both approaches are
yielding diminishing returns:</t> yielding diminishing returns:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>Recent state-of-the-art active queue management (AQM) in the <li>Recent state-of-the-art AQM in the
network, e.g. FQ-CoDel <xref target="RFC8290" format="default"/>, network, e.g., Flow Queue CoDel <xref target="RFC8290" format="defau
PIE <xref target="RFC8033" format="default"/>, Adaptive RED <xref ta lt"/>,
rget="ARED01" format="default"/> ) has reduced queuing delay for all traffic, no Proportional Integral controller Enhanced (PIE) <xref target="RFC803
t 3" format="default"/>, and Adaptive Random Early Detection (ARED) <xref target="
ARED01" format="default"/>), has reduced queuing delay for all traffic, not
just a select few applications. However, no matter how good the just a select few applications. However, no matter how good the
AQM, the capacity-seeking (sawtoothing) rate of TCP-like AQM, the capacity-seeking (sawtoothing) rate of TCP-like
congestion controls represents a lower limit that will either congestion controls represents a lower limit that will cause either
cause queuing delay to vary or cause the link to be the queuing delay to vary or the link to be
under-utilized. These AQMs are tuned to allow a typical underutilized.
capacity-seeking Reno-friendly flow to induce an average queue These AQMs are tuned to allow a typical
that roughly doubles the base RTT, adding 5-15 ms of queuing on capacity-seeking TCP-Reno-friendly flow to induce an average queue
average (cf. 500 microseconds with L4S for the same mix of that roughly doubles the base round-trip time (RTT), adding 5-15 ms
long-running and web traffic). However, for many applications low of queuing on
average for a mix of long-running flows and web traffic (cf. 500 mic
roseconds with L4S for the same traffic mix <xref target="L4Seval22" format="def
ault"/>). However, for many applications, low
delay is not useful unless it is consistently low. With these delay is not useful unless it is consistently low. With these
AQMs, 99th percentile queuing delay is 20-30 ms (cf. 2 ms with the AQMs, 99th percentile queuing delay is 20-30 ms (cf. 2 ms with the
same traffic over L4S).</li> same traffic over L4S).</li>
<li>Similarly, recent research into using e2e congestion control
without needing an AQM in the network (e.g. BBR <xref target="I-D.ca <li>Similarly, recent research into using end-to-end congestion contro
rdwell-iccrg-bbr-congestion-control" format="default"/>) seems to l
have hit a similar lower limit to queuing delay of about 20ms on without needing an AQM in the network (e.g., Bottleneck Bandwidth an
average, but there are also regular 25ms delay spikes due to d Round-trip propagation time (BBR) <xref target="I-D.cardwell-iccrg-bbr-conges
bandwidth probes and 60ms spikes due to flow-starts.</li> tion-control" format="default"/>) seems to
have hit a similar queuing delay floor of about 20 ms on
average, but there are also regular 25 ms delay spikes due to
bandwidth probes and 60 ms spikes due to flow-starts.</li>
</ul> </ul>
<t>L4S learns from the experience of Data Center TCP <xref target="RFC82 <t>L4S learns from the experience of Data Center TCP (DCTCP) <xref targe
57" format="default"/>, which shows the power of complementary changes t="RFC8257" format="default"/>, which shows the power of complementary changes
both in the network and on end-systems. DCTCP teaches us that two both in the network and on end systems. DCTCP teaches us that two
small but radical changes to congestion control are needed to cut the small but radical changes to congestion control are needed to cut the
two major outstanding causes of queuing delay variability:</t> two major outstanding causes of queuing delay variability:</t>
<ol spacing="normal" type="1"><li>Far smaller rate variations (sawteeth) than Reno-friendly <ol spacing="normal" type="1"><li>Far smaller rate variations (sawteeth) than Reno-friendly
congestion controls;</li> congestion controls.</li>
<li>A shift of smoothing and hence smoothing delay from network to <li>A shift of smoothing and hence smoothing delay from network to
sender.</li> sender.</li>
</ol> </ol>
<t>Without the former, a 'Classic' (e.g. Reno-friendly) <t>Without the former, a 'Classic' (e.g., Reno-friendly)
flow's round trip time (RTT) varies between roughly 1 and 2 times the flow's RTT varies between roughly 1 and 2 times the
base RTT between the machines in question. Without the latter a base RTT between the machines in question. Without the latter, a
'Classic' flow's response to changing events is delayed by a 'Classic' flow's response to changing events is delayed by a
worst-case (transcontinental) RTT, which could be hundreds of times worst-case (transcontinental) RTT, which could be hundreds of times
the actual smoothing delay needed for the RTT of typical traffic from the actual smoothing delay needed for the RTT of typical traffic from
localized CDNs.</t> localized Content Delivery Networks (CDNs).</t>
<t>These changes are the two main features of the family of so-called <t>These changes are the two main features of the family of so-called
'Scalable' congestion controls (which includes DCTCP, TCP Prague and 'Scalable' congestion controls (which include DCTCP, Prague, and
SCReAM). Both these changes only reduce delay in combination with a Self-Clocked Rate Adaptation for Multimedia (SCReAM)). Both of these cha
complementary change in the network and they are both only feasible nges only reduce delay in combination with a
complementary change in the network, and they are both only feasible
with ECN, not drop, for the signalling:</t> with ECN, not drop, for the signalling:</t>
<ol spacing="normal" type="1"><li>The smaller sawteeth allow an extremel <ol spacing="normal">
y shallow ECN <li>The smaller sawteeth allow an extremely shallow ECN
packet-marking threshold in the queue.</li> packet-marking threshold in the queue.</li>
<li>And no smoothing in the network means that every fluctuation of <li>No smoothing in the network means that every fluctuation of
the queue is signalled immediately.</li> the queue is signalled immediately.</li>
</ol> </ol>
<t>Without ECN, either of these would lead to very high loss <t>Without ECN, either of these would lead to very high loss
levels. But, with ECN, the resulting high marking levels are just levels. In contrast, with ECN, the resulting high marking levels are jus
signals, not impairments. (Note that BBRv2 <xref target="BBRv2" format=" t
default"/> signals, not impairments.
combines the best of both worlds - it works as a scalable congestion (Note that BBRv2 <xref target="BBRv2" format="default"/>
control when ECN is available, but also aims to minimize delay when it combines the best of both worlds -- it works as a Scalable congestion
isn't.)</t> control when ECN is available, but it also aims to minimize delay when E
CN
is absent.)</t>
<t>However, until now, Scalable congestion controls (like DCTCP) did <t>However, until now, Scalable congestion controls (like DCTCP) did
not co-exist well in a shared ECN-capable queue with existing Classic not coexist well in a shared ECN-capable queue with existing Classic
(e.g. Reno <xref target="RFC5681" format="default"/> or Cubic <xref targ (e.g., Reno <xref target="RFC5681" format="default"/> or CUBIC <xref tar
et="RFC8312" format="default"/>) congestion controls -- Scalable controls are get="RFC8312" format="default"/>) congestion controls -- Scalable controls are
so aggressive that these 'Classic' algorithms would drive themselves so aggressive that these 'Classic' algorithms would drive themselves
to a small capacity share. Therefore, until now, L4S controls could to a small capacity share. Therefore, until now, L4S controls could
only be deployed where a clean-slate environment could be arranged, only be deployed where a clean-slate environment could be arranged,
such as in private data centres (hence the name DCTCP).</t> such as in private data centres (hence the name DCTCP).</t>
<t>One way to solve the problem of coexistence between Scalable and <t>One way to solve the problem of coexistence between Scalable and
Classic flows is to use a per-flow-queuing approach such as Classic flows is to use a per-flow-queuing (FQ) approach such as
FQ-CoDel <xref target="RFC8290" format="default"/>. It classifies packet FQ-CoDel <xref target="RFC8290" format="default"/>. It classifies packet
s by flow s by flow
identifier into separate queues in order to isolate sparse flows from identifier into separate queues in order to isolate sparse flows from
the higher latency in the queues assigned to heavier flows. However, the higher latency in the queues assigned to heavier flows. However,
if a Classic flow needs both low delay and high throughput, having a if a Classic flow needs both low delay and high throughput, having a
queue to itself does not isolate it from the harm it causes to itself. queue to itself does not isolate it from the harm it causes to itself.
Also FQ approaches need to inspect flow identifiers, which is not Also FQ approaches need to inspect flow identifiers, which is not
always practical.</t> always practical.</t>
<t>In summary, Scalable congestion controls address the root cause of <t>In summary, Scalable congestion controls address the root cause of
the latency, loss and scaling problems with Classic congestion the latency, loss and scaling problems with Classic congestion
controls. Both FQ and DualQ AQMs can be enablers for this smooth low controls. Both FQ and DualQ AQMs can be enablers for this smooth low-lat
latency scalable behaviour. The DualQ approach is particularly useful ency
scalable behaviour. The DualQ approach is particularly useful
because identifying flows is sometimes not practical or desirable.</t> because identifying flows is sometimes not practical or desirable.</t>
</section> </section>
<section anchor="dualq_scope" numbered="true" toc="default"> <section anchor="dualq_scope" numbered="true" toc="default">
<name>Context, Scope &amp; Applicability</name> <name>Context, Scope, and Applicability</name>
<t>L4S involves complementary changes in the network and on <t>L4S involves complementary changes in the network and on
end-systems:</t> end systems:</t>
<dl newline="false" spacing="normal"> <dl newline="true" spacing="normal">
<dt>Network:</dt> <dt>Network:</dt>
<dd>A DualQ Coupled AQM (defined in the present <dd>A DualQ Coupled AQM (defined in the present
document) or a modification to flow-queue AQMs (described in document) or a modification to flow queue AQMs (described in paragra
section 4.2.b of the L4S architecture <xref target="I-D.ietf-tsvwg-l ph "b" in
4s-arch" format="default"/>);</dd> Section <xref target="RFC9330" sectionFormat="bare" section="4.2"/> o
<dt>End-system:</dt> f the L4S architecture <xref target="RFC9330" format="default"/>).</dd>
<dd>A Scalable congestion control (defined <dt>End system:</dt>
in section 4 of the L4S ECN protocol <xref target="I-D.ietf-tsvwg-ec <dd>A Scalable congestion control (defined in Section <xref target="RF
n-l4s-id" format="default"/>).</dd> C9331" sectionFormat="bare" section="4"/> of the L4S ECN protocol spec <xref tar
get="RFC9331" format="default"/>).</dd>
<dt>Packet identifier:</dt> <dt>Packet identifier:</dt>
<dd>The network and end-system parts <dd>The network and end-system parts
of L4S can be deployed incrementally, because they both identify of L4S can be deployed incrementally, because they both identify
L4S packets using the experimentally assigned explicit congestion L4S packets using the experimentally assigned ECN codepoints in the
notification (ECN) codepoints in the IP header: ECT(1) and IP header: ECT(1) and
CE <xref target="RFC8311" format="default"/> <xref target="I-D.ietf- CE <xref target="RFC8311" format="default"/> <xref target="RFC9331"
tsvwg-ecn-l4s-id" format="default"/>.</dd> format="default"/>.</dd>
</dl> </dl>
<t>Data Center TCP (DCTCP <xref target="RFC8257" format="default"/>) is an example <t>DCTCP <xref target="RFC8257" format="default"/> is an example
of a Scalable congestion control for controlled environments that has of a Scalable congestion control for controlled environments that has
been deployed for some time in Linux, Windows and FreeBSD operating been deployed for some time in Linux, Windows, and FreeBSD operating
systems. During the progress of this document through the IETF a systems. During the progress of this document through the IETF, a
number of other Scalable congestion controls were implemented, number of other Scalable congestion controls were implemented,
e.g. TCP Prague <xref target="I-D.briscoe-iccrg-prague-congestion-contro e.g., Prague over TCP and QUIC <xref target="I-D.briscoe-iccrg-prague-co
l" format="default"/> <xref target="PragueLinux" format="default"/>, BBRv2 <xref ngestion-control" format="default"/> <xref target="PragueLinux" format="default"
target="BBRv2" format="default"/>, <xref target="I-D.cardwell-iccrg-bbr-congest />, BBRv2 <xref target="BBRv2" format="default"/> <xref target="I-D.cardwell-icc
ion-control" format="default"/>, QUIC Prague and rg-bbr-congestion-control" format="default"/>, and
the L4S variant of SCREAM for real-time media <xref target="RFC8298" for the L4S variant of SCReAM for real-time media <xref target="SCReAM-L4S"
mat="default"/>.</t> format="default"/> <xref target="RFC8298" format="default"/>.</t>
<t>The focus of this specification is to enable deployment of the
<t>The focus of this specification is to enable deployment of the
network part of the L4S service. Then, without any management network part of the L4S service. Then, without any management
intervention, applications can exploit this new network capability as intervention, applications can exploit this new network capability as
their operating systems migrate to Scalable congestion controls, which the applications or their operating systems migrate to Scalable congesti on controls, which
can then evolve <em>while</em> their benefits are can then evolve <em>while</em> their benefits are
being enjoyed by everyone on the Internet.</t> being enjoyed by everyone on the Internet.</t>
<t>The DualQ Coupled AQM framework can incorporate any AQM designed <t>The DualQ Coupled AQM framework can incorporate any AQM designed
for a single queue that generates a statistical or deterministic for a single queue that generates a statistical or deterministic
mark/drop probability driven by the queue dynamics. Pseudocode mark/drop probability driven by the queue dynamics. Pseudocode
examples of two different DualQ Coupled AQMs are given in the examples of two different DualQ Coupled AQMs are given in the
appendices. In many cases the framework simplifies the basic control appendices.
algorithm, and requires little extra processing. Therefore, it is In many cases the framework simplifies the basic control
algorithm and requires little extra processing.
Therefore, it is
believed the Coupled AQM would be applicable and easy to deploy in all believed the Coupled AQM would be applicable and easy to deploy in all
types of buffers; buffers in cost-reduced mass-market residential types of buffers such as buffers in cost-reduced mass-market residential
equipment; buffers in end-system stacks; buffers in carrier-scale equipment; buffers in end-system stacks; buffers in carrier-scale
equipment including remote access servers, routers, firewalls and equipment including remote access servers, routers, firewalls, and
Ethernet switches; buffers in network interface cards, buffers in Ethernet switches; buffers in network interface cards; buffers in
virtualized network appliances, hypervisors, and so on.</t> virtualized network appliances, hypervisors; and so on.</t>
<t>For the public Internet, nearly all the benefit will typically be <t>For the public Internet, nearly all the benefit will typically be
achieved by deploying the Coupled AQM into either end of the access achieved by deploying the Coupled AQM into either end of the access
link between a 'site' and the Internet, which is invariably the link between a 'site' and the Internet, which is invariably the
bottleneck (see section 6.4 of<xref target="I-D.ietf-tsvwg-l4s-arch" for mat="default"/> bottleneck (see <xref target="RFC9330" sectionFormat="of" section="6.4"/ >
about deployment, which also defines the term 'site' to mean a home, about deployment, which also defines the term 'site' to mean a home,
an office, a campus or mobile user equipment).</t> an office, a campus, or mobile user equipment).</t>
<t>Latency is not the only concern of L4S:</t> <t>Latency is not the only concern of L4S:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>The "Low Loss" part of the name denotes that L4S generally <li>The 'Low Loss' part of the name denotes that L4S generally
achieves zero congestion loss (which would otherwise cause achieves zero congestion loss (which would otherwise cause
retransmission delays), due to its use of ECN.</li> retransmission delays), due to its use of ECN.</li>
<li>The "Scalable throughput" part of the name denotes that the <li>The 'Scalable throughput' part of the name denotes that the
per-flow throughput of Scalable congestion controls should scale per-flow throughput of Scalable congestion controls should scale
indefinitely, avoiding the imminent scaling problems with indefinitely, avoiding the imminent scaling problems with
'TCP-Friendly' congestion control algorithms <xref target="RFC3649" format="default"/>.</li> 'TCP-Friendly' congestion control algorithms <xref target="RFC3649" format="default"/>.</li>
</ul> </ul>
<t>The former is clearly in scope of this AQM document. However, <t>The former is clearly in scope of this AQM document. However,
the latter is an outcome of the end-system behaviour, and therefore the latter is an outcome of the end-system behaviour and is therefore
outside the scope of this AQM document, even though the AQM is an outside the scope of this AQM document, even though the AQM is an
enabler.</t> enabler.</t>
<t>The overall L4S architecture <xref target="I-D.ietf-tsvwg-l4s-arch" f ormat="default"/> gives more detail, including on <t>The overall L4S architecture <xref target="RFC9330" format="default"/ > gives more detail, including on
wider deployment aspects such as backwards compatibility of Scalable wider deployment aspects such as backwards compatibility of Scalable
congestion controls in bottlenecks where a DualQ Coupled AQM has not congestion controls in bottlenecks where a DualQ Coupled AQM has not
been deployed. The supporting papers <xref target="DualPI2Linux" format= been deployed. The supporting papers <xref target="L4Seval22"/>, <xref t
"default"/>, arget="DualPI2Linux" format="default"/>,
<xref target="PI2" format="default"/>, <xref target="DCttH19" format="de <xref target="PI2" format="default"/>, and <xref target="PI2param" forma
fault"/> and <xref target="PI2param" format="default"/> give the full rationale t="default"/> give the full rationale for the AQM design, both
for the AQM's design, both
discursively and in more precise mathematical form, as well as the discursively and in more precise mathematical form, as well as the
results of performance evaluations. The main results have been results of performance evaluations. The main results have been
validated independently when using the Prague congestion control <xref t arget="Boru20" format="default"/> (experiments are run using Prague and DCTCP, b ut validated independently when using the Prague congestion control <xref t arget="Boru20" format="default"/> (experiments are run using Prague and DCTCP, b ut
only the former are relevant for validation, because Prague fixes a only the former is relevant for validation, because Prague fixes a
number of problems with the Linux DCTCP code that make it unsuitable number of problems with the Linux DCTCP code that make it unsuitable
for the public Internet).</t> for the public Internet).</t>
</section> </section>
<section anchor="dualq_Terminology" numbered="true" toc="default"> <section anchor="dualq_Terminology" numbered="true" toc="default">
<name>Terminology</name> <name>Terminology</name>
<t>The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", <t>
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this The key words "<bcp14>MUST</bcp14>", "<bcp14>MUST NOT</bcp14>", "<bcp14>REQU
document are to be interpreted as described in <xref target="RFC2119" fo IRED</bcp14>", "<bcp14>SHALL</bcp14>", "<bcp14>SHALL
rmat="default"/> <xref target="RFC8174" format="default"/> when, and only when, NOT</bcp14>", "<bcp14>SHOULD</bcp14>", "<bcp14>SHOULD NOT</bcp14>", "<bcp14>
they RECOMMENDED</bcp14>", "<bcp14>NOT RECOMMENDED</bcp14>",
appear in all capitals, as shown here.</t> "<bcp14>MAY</bcp14>", and "<bcp14>OPTIONAL</bcp14>" in this document are to
<t>The DualQ Coupled AQM uses two queues for two services. Each of the be interpreted as
following terms identifies both the service and the queue that described in BCP&nbsp;14 <xref target="RFC2119"/> <xref target="RFC8174"/>
provides the service:</t> when, and only when, they appear in all capitals, as shown here.
</t>
<t>The DualQ Coupled AQM uses two queues for two services:</t>
<dl newline="false" spacing="normal"> <dl newline="false" spacing="normal">
<dt>Classic service/queue:</dt> <dt>Classic Service/Queue:</dt>
<dd>The Classic service is <dd>The Classic service is
intended for all the congestion control behaviours that co-exist intended for all the congestion control behaviours that coexist
with Reno <xref target="RFC5681" format="default"/> (e.g. Reno itsel with Reno <xref target="RFC5681" format="default"/> (e.g., Reno itse
f, lf,
Cubic <xref target="RFC8312" format="default"/>, TFRC <xref target=" CUBIC <xref target="RFC8312" format="default"/>, and TFRC <xref targ
RFC5348" format="default"/>).</dd> et="RFC5348" format="default"/>). The term 'Classic queue' means a queue providi
<dt>Low-Latency, Low-Loss Scalable throughput (L4S) service/queue:</dt ng the Classic service.</dd>
>
<dt>Low Latency, Low Loss, and Scalable throughput (L4S) Service/Queue
:</dt>
<dd>The <dd>The
'L4S' service is intended for traffic from scalable congestion 'L4S' service is intended for traffic from Scalable congestion
control algorithms, such as TCP Prague <xref target="I-D.briscoe-icc control algorithms, such as the Prague congestion control <xref targ
rg-prague-congestion-control" format="default"/>, which was et="I-D.briscoe-iccrg-prague-congestion-control" format="default"/>, which was
derived from Data Center TCP <xref target="RFC8257" format="default" derived from Data Center TCP <xref target="RFC8257" format="default"
/>. The />. The
L4S service is for more general traffic than just TCP Prague L4S service is for more general traffic than just Prague
-- it allows the set of congestion controls with similar -- it allows the set of congestion controls with similar
scaling properties to Prague to evolve, such as the examples of scaling properties to Prague to evolve, such as the examples listed
Scalable congestion controls listed below (Relentless, SCReAM, below (Relentless, SCReAM, etc.). The term 'L4S queue' means a queue providing t
etc.).</dd> he L4S service.</dd>
<dt>Classic Congestion Control:</dt> <dt>Classic Congestion Control:</dt>
<dd>A congestion control <dd>A congestion control
behaviour that can co-exist with standard TCP Reno <xref target="RFC behaviour that can coexist with standard Reno <xref target="RFC5681"
5681" format="default"/> without causing significantly negative impact format="default"/> without causing significantly negative impact
on its flow rate <xref target="RFC5033" format="default"/>. With Cla on its flow rate <xref target="RFC5033" format="default"/>. With Cla
ssic ssic
congestion controls, such as Reno or Cubic, because flow rate has congestion controls, such as Reno or CUBIC, because flow rate has
scaled since TCP congestion control was first designed in 1988, it scaled since TCP congestion control was first designed in 1988, it
now takes hundreds of round trips (and growing) to recover after a now takes hundreds of round trips (and growing) to recover after a
congestion signal (whether a loss or an ECN mark) as shown in the congestion signal (whether a loss or an ECN mark) as shown in the
examples in section 5.1 of the L4S architecture <xref target="I-D.ie examples in Section <xref target="RFC9330" sectionFormat="bare" sect
tf-tsvwg-l4s-arch" format="default"/> and in <xref target="RFC3649" format="defa ion="5.1"/> of the L4S architecture <xref target="RFC9330"/> and in <xref target
ult"/>. Therefore, control of queuing and utilization ="RFC3649" format="default"/>. Therefore, control of queuing and utilization
becomes very slack, and the slightest disturbances (e.g. from becomes very slack, and the slightest disturbances (e.g., from
new flows starting) prevent a high rate from being attained.</dd> new flows starting) prevent a high rate from being attained.</dd>
<dt>Scalable Congestion Control:</dt> <dt>Scalable Congestion Control:</dt>
<dd>A congestion control <dd>A congestion control
where the average time from one congestion signal to the next (the where the average time from one congestion signal to the next (the
recovery time) remains invariant as the flow rate scales, all recovery time) remains invariant as flow rate scales, all
other factors being equal. This maintains the same degree of other factors being equal. This maintains the same degree of
control over queueing and utilization whatever the flow rate, as control over queuing and utilization whatever the flow rate, as
well as ensuring that high throughput is robust to disturbances. well as ensuring that high throughput is robust to disturbances.
For instance, DCTCP averages 2 congestion signals per round-trip For instance, DCTCP averages 2 congestion signals per round trip,
whatever the flow rate, as do other recently developed scalable whatever the flow rate, as do other recently developed Scalable
congestion controls, e.g. Relentless TCP <xref target="I-D.mathis-ic congestion controls, e.g., Relentless TCP <xref target="I-D.mathis-i
crg-relentless-tcp" format="default"/>, TCP Prague <xref target="I-D.briscoe-icc ccrg-relentless-tcp" format="default"/>, Prague <xref target="I-D.briscoe-iccrg-
rg-prague-congestion-control" format="default"/>, <xref target="PragueLinux" for prague-congestion-control" format="default"/> <xref target="PragueLinux" format=
mat="default"/>, BBRv2 <xref target="BBRv2" format="default"/>, <xref target="I- "default"/>, BBRv2 <xref target="BBRv2" format="default"/> <xref target="I-D.car
D.cardwell-iccrg-bbr-congestion-control" format="default"/> and the L4S dwell-iccrg-bbr-congestion-control" format="default"/>, and the L4S
variant of SCREAM for real-time media <xref target="SCReAM" format=" variant of SCReAM for real-time media <xref target="SCReAM-L4S" form
default"/>, <xref target="RFC8298" format="default"/>). For the public at="default"/> <xref target="RFC8298" format="default"/>. For the public
Internet a Scalable transport has to comply with the requirements Internet, a Scalable transport has to comply with the requirements
in Section 4 of <xref target="I-D.ietf-tsvwg-ecn-l4s-id" format="def in <xref target="RFC9331" sectionFormat="of" section="4"/> (a.k.a. t
ault"/> he 'Prague L4S requirements').</dd>
(aka. the 'Prague L4S requirements').</dd>
<dt>C:</dt> <dt>C:</dt>
<dd>Abbreviation for Classic, e.g. when used as <dd>Abbreviation for Classic, e.g., when used as
a subscript.</dd> a subscript.</dd>
<dt>L:</dt> <dt>L:</dt>
<dd> <dd>
<t>Abbreviation for L4S, e.g. when used as a <t>Abbreviation for L4S, e.g., when used as a
subscript.</t> subscript.</t>
<t>The terms Classic or L4S can <t>The terms Classic or L4S can
also qualify other nouns, such as 'codepoint', 'identifier', also qualify other nouns, such as 'codepoint', 'identifier',
'classification', 'packet', 'flow'. For example: an L4S packet 'classification', 'packet', and 'flow'. For example, an L4S packet
means a packet with an L4S identifier sent from an L4S congestion means a packet with an L4S identifier sent from an L4S congestion
control.</t> control.</t>
<t>Both Classic and L4S services can <t>Both Classic and L4S services can
cope with a proportion of unresponsive or less-responsive traffic cope with a proportion of unresponsive or less-responsive traffic
as well, but in the L4S case its rate has to be smooth enough or as well but, in the L4S case, its rate has to be smooth enough or
low enough not to build a queue (e.g. DNS, VoIP, game sync low enough to not build a queue (e.g., DNS, Voice over IP (VoIP), ga
me sync
datagrams, etc.). The DualQ Coupled AQM behaviour is defined to be datagrams, etc.). The DualQ Coupled AQM behaviour is defined to be
similar to a single FIFO queue with respect to unresponsive and similar to a single First-In, First-Out (FIFO) queue with respect to unresponsive and
overload traffic.</t> overload traffic.</t>
</dd> </dd>
<dt>Reno-friendly:</dt> <dt>Reno-friendly:</dt>
<dd>The subset of Classic traffic that is <dd>The subset of Classic traffic that is
friendly to the standard Reno congestion control defined for TCP friendly to the standard Reno congestion control defined for TCP
in <xref target="RFC5681" format="default"/>. Reno-friendly is used in <xref target="RFC5681" format="default"/>.
in place of The TFRC spec <xref target="RFC5348"/> indirectly implies that 'friendly' is
defined as "generally within a factor of two of the sending rate
of a TCP flow under the same conditions". 'Reno-friendly' is used here in place
of
'TCP-friendly', given the latter has become imprecise, because the 'TCP-friendly', given the latter has become imprecise, because the
TCP protocol is now used with so many different congestion control TCP protocol is now used with so many different congestion control
behaviours, and Reno is used in non-TCP transports such as behaviours, and Reno is used in non-TCP transports, such as
QUIC.</dd> QUIC <xref target="RFC9000"/>.</dd>
<dt>DualQ or DualQ AQM:</dt>
<dd>Used loosely as shorthand for a Dual-Queue Coupled AQM, where the
context
makes 'Coupled AQM' obvious.</dd>
<dt>Classic ECN:</dt> <dt>Classic ECN:</dt>
<dd> <dd>
<t>The original Explicit Congestion <t>The original Explicit Congestion
Notification (ECN) protocol <xref target="RFC3168" format="default"/ Notification (ECN) protocol <xref target="RFC3168" format="default"/
>, which > that
requires ECN signals to be treated the same as drops, both when requires ECN signals to be treated as equivalent to drops, both when
generated in the network and when responded to by the generated in the network and when responded to by the
sender.</t> sender.</t>
<t>For L4S, the names used for the <t>For L4S, the names used for the four codepoints of the 2-bit IP-E
four codepoints of the 2-bit IP-ECN field are unchanged from those CN field are unchanged from those
defined in <xref target="RFC3168" format="default"/>: Not ECT, ECT(0 defined in the ECN spec <xref target="RFC3168" format="default"/>, i
), ECT(1) and .e., Not-ECT, ECT(0), ECT(1), and
CE, where ECT stands for ECN-Capable Transport and CE stands for CE, where ECT stands for ECN-Capable Transport and CE stands for
Congestion Experienced. A packet marked with the CE codepoint is Congestion Experienced. A packet marked with the CE codepoint is
termed 'ECN-marked' or sometimes just 'marked' where the context termed 'ECN-marked' or sometimes just 'marked' where the context
makes ECN obvious.</t> makes ECN obvious.</t>
</dd> </dd>
</dl> </dl>
</section> </section>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Features</name> <name>Features</name>
<t>The AQM couples marking and/or dropping from the Classic queue to <t>The AQM couples marking and/or dropping from the Classic queue to
the L4S queue in such a way that a flow will get roughly the same the L4S queue in such a way that a flow will get roughly the same
throughput whichever it uses. Therefore, both queues can feed into the throughput whichever it uses. Therefore, both queues can feed into the
full capacity of a link and no rates need to be configured for the full capacity of a link, and no rates need to be configured for the
queues. The L4S queue enables Scalable congestion controls like DCTCP queues.
or TCP Prague to give very low and predictably low latency, without The L4S queue enables Scalable congestion controls like DCTCP
or Prague to give very low and consistently low latency, without
compromising the performance of competing 'Classic' Internet compromising the performance of competing 'Classic' Internet
traffic.</t> traffic.</t>
<t>Thousands of tests have been conducted in a typical fixed <t>Thousands of tests have been conducted in a typical fixed
residential broadband setting. Experiments used a range of base round residential broadband setting. Experiments used a range of base round-tr
trip delays up to 100ms and link rates up to 200 Mb/s between the data ip
delays up to 100 ms and link rates up to 200 Mb/s between the data
centre and home network, with varying amounts of background traffic in centre and home network, with varying amounts of background traffic in
both queues. For every L4S packet, the AQM kept the average queuing both queues. For every L4S packet, the AQM kept the average queuing
delay below 1ms (or 2 packets where serialization delay exceeded 1ms delay below 1 ms (or 2 packets where serialization delay exceeded 1 ms
on slower links), with 99th percentile no worse than 2ms. No losses at on slower links), with the 99th percentile being no worse than 2 ms. No
losses at
all were introduced by the L4S AQM. Details of the extensive all were introduced by the L4S AQM. Details of the extensive
experiments are available <xref target="DualPI2Linux" format="default"/> experiments are available in <xref target="L4Seval22" format="default"/>
, <xref target="PI2" format="default"/>, <xref target="DCttH19" format="default" and <xref target="DualPI2Linux" format="default"/>.
/>. Subjective testing using Subjective testing using
very demanding high bandwidth low latency applications over a single very demanding high-bandwidth low-latency applications over a single
shared access link is also described in <xref target="L4Sdemo16" format= shared access link is also described in <xref target="L4Sdemo16" format=
"default"/> and summarized in the section about applications "default"/> and summarized in Section <xref
in the L4S architecture <xref target="I-D.ietf-tsvwg-l4s-arch" format="d target="RFC9330" sectionFormat="bare" section="6.1"/> of the L4S archite
efault"/> cture <xref target="RFC9330" format="default"/>.
.</t> </t>
<t>In all these experiments, the host was connected to the home <t>In all these experiments, the host was connected to the home
network by fixed Ethernet, in order to quantify the queuing delay that network by fixed Ethernet, in order to quantify the queuing delay that
can be achieved by a user who cares about delay. It should be can be achieved by a user who cares about delay. It should be
emphasized that L4S support at the bottleneck link cannot 'undelay' emphasized that L4S support at the bottleneck link cannot 'undelay'
bursts introduced by another link on the path, for instance by legacy bursts introduced by another link on the path, for instance by legacy
Wi-Fi equipment. However, if L4S support is added to the queue feeding Wi-Fi equipment. However, if L4S support is added to the queue feeding
the <em>outgoing</em> WAN link of a home gateway, the <em>outgoing</em> WAN link of a home gateway,
it would be counterproductive not to also reduce the burstiness of the it would be counterproductive not to also reduce the burstiness of the
<em>incoming</em> Wi-Fi. Also, trials of Wi-Fi <em>incoming</em> Wi-Fi. Also, trials of Wi-Fi
equipment with an L4S DualQ Coupled AQM on the <em>outgoing</em> equipment with an L4S DualQ Coupled AQM on the <em>outgoing</em>
Wi-Fi interface are in progress, and early results of an L4S DualQ Wi-Fi interface are in progress, and early results of an L4S DualQ
Coupled AQM in a 5G radio access network testbed with emulated outdoor Coupled AQM in a 5G radio access network testbed with emulated outdoor
cell edge radio fading are given in <xref target="L4S_5G" format="defaul t"/>.</t> cell edge radio fading are given in <xref target="L4S_5G" format="defaul t"/>.</t>
<t>Unlike Diffserv Expedited Forwarding, the L4S queue does not have <t>Unlike Diffserv EF, the L4S queue does not have
to be limited to a small proportion of the link capacity in order to to be limited to a small proportion of the link capacity in order to
achieve low delay. The L4S queue can be filled with a heavy load of achieve low delay. The L4S queue can be filled with a heavy load of
capacity-seeking flows (TCP Prague etc.) and still achieve low delay. capacity-seeking flows (Prague, BBRv2, etc.) and still achieve low delay .
The L4S queue does not rely on the presence of other traffic in the The L4S queue does not rely on the presence of other traffic in the
Classic queue that can be 'overtaken'. It gives low latency to L4S Classic queue that can be 'overtaken'.
It gives low latency to L4S
traffic whether or not there is Classic traffic. The tail latency of traffic whether or not there is Classic traffic. The tail latency of
traffic served by the Classic AQM is sometimes a little better traffic served by the Classic AQM is sometimes a little better,
sometimes a little worse, when a proportion of the traffic is L4S.</t> sometimes a little worse, when a proportion of the traffic is L4S.</t>
<t>The two queues are only necessary because:</t> <t>The two queues are only necessary because:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>the large variations (sawteeth) of Classic flows need roughly a <li>The large variations (sawteeth) of Classic flows need roughly a
base RTT of queuing delay to ensure full utilization</li> base RTT of queuing delay to ensure full utilization.</li>
<li>Scalable flows do not need a queue to keep utilization high, <li>Scalable flows do not need a queue to keep utilization high,
but they cannot keep latency predictably low if they are mixed but they cannot keep latency consistently low if they are mixed
with Classic traffic,</li> with Classic traffic.</li>
</ul> </ul>
<t>The L4S queue has latency priority within sub-round trip <t>The L4S queue has latency priority within sub-round-trip
timescales, but over longer periods the coupling from the Classic to timescales, but over longer periods the coupling from the Classic to
the L4S AQM (explained below) ensures that it does not have bandwidth the L4S AQM (explained below) ensures that it does not have bandwidth
priority over the Classic queue.</t> priority over the Classic queue.</t>
</section> </section>
</section> </section>
<section anchor="dualq_algo" numbered="true" toc="default"> <section anchor="dualq_algo" numbered="true" toc="default">
<name>DualQ Coupled AQM</name> <name>DualQ Coupled AQM</name>
<t>There are two main aspects to the approach:</t> <t>There are two main aspects to the DualQ Coupled AQM approach:</t>
<ul spacing="normal"> <ol spacing="normal">
<li>The Coupled AQM that addresses throughput equivalence between <li>The Coupled AQM that addresses throughput equivalence between
Classic (e.g. Reno, Cubic) flows and L4S flows (that satisfy Classic (e.g., Reno or CUBIC) flows and L4S flows (that satisfy
the Prague L4S requirements).</li> the Prague L4S requirements).</li>
<li>The Dual Queue structure that provides latency separation for L4S <li>The Dual-Queue structure that provides latency separation for L4S
flows to isolate them from the typically large Classic queue.</li> flows to isolate them from the typically large Classic queue.</li>
</ul> </ol>
<!--<t>The following subsections descrbe these two aspects, and how
packets are classified between the two queues, then a likely overall
structure of a DualQ Coupled AQM is given. The present document applies
irrespective of which particular AQMs are used for each queue. So,
although the structure is intended to be generic, it might not fit well
around types of AQM yet to be considered. Finally normative requirements
are given that apply to any specific DualQ Coupled AQM implementation,
irrespective of which AQMs it uses. Pseudocode of specific examples are
given in non-normative appendices.</t>-->
<section anchor="dualq_coupled" numbered="true" toc="default"> <section anchor="dualq_coupled" numbered="true" toc="default">
<name>Coupled AQM</name> <name>Coupled AQM</name>
<t>In the 1990s, the `TCP formula' was derived for the relationship <t>In the 1990s, the 'TCP formula' was derived for the relationship
between the steady-state congestion window, cwnd, and the drop between the steady-state congestion window, cwnd, and the drop
probability, p of standard Reno congestion control <xref target="RFC5681 " format="default"/>. To a first order approximation, the steady-state probability, p of standard Reno congestion control <xref target="RFC5681 " format="default"/>. To a first-order approximation, the steady-state
cwnd of Reno is inversely proportional to the square root of p.</t> cwnd of Reno is inversely proportional to the square root of p.</t>
<t>The design focuses on Reno as the worst case, because if it does no <t>The design focuses on Reno as the worst case, because if it does no
harm to Reno, it will not harm Cubic or any traffic designed to be harm to Reno, it will not harm CUBIC or any traffic designed to be
friendly to Reno. TCP Cubic implements a Reno-compatibility mode, friendly to Reno. TCP CUBIC implements a Reno-friendly mode,
which is relevant for typical RTTs under 20ms as long as the which is relevant for typical RTTs under 20 ms as long as the
throughput of a single flow is less than about 350Mb/s. In such cases throughput of a single flow is less than about 350 Mb/s. In such cases,
it can be assumed that Cubic traffic behaves similarly to Reno. The it can be assumed that CUBIC traffic behaves similarly to Reno. The
term 'Classic' will be used for the collection of Reno-friendly term 'Classic' will be used for the collection of Reno-friendly
traffic including Cubic and potentially other experimental congestion traffic including CUBIC and potentially other experimental congestion
controls intended not to significantly impact the flow rate of controls intended not to significantly impact the flow rate of
Reno.</t> Reno.</t>
<t>A supporting paper <xref target="PI2" format="default"/> includes the <t>A supporting paper <xref target="PI2" format="default"/> includes the
derivation of the equivalent rate equation for DCTCP, for which cwnd derivation of the equivalent rate equation for DCTCP, for which cwnd
is inversely proportional to p (not the square root), where in this is inversely proportional to p (not the square root), where in this
case p is the ECN marking probability. DCTCP is not the only case p is the ECN-marking probability. DCTCP is not the only
congestion control that behaves like this, so the term 'Scalable' will congestion control that behaves like this, so the term 'Scalable' will
be used for all similar congestion control behaviours (see examples in be used for all similar congestion control behaviours (see examples in
<xref target="dualq_scope" format="default"/>). The term 'L4S' is used f or traffic <xref target="dualq_scope" format="default"/>). The term 'L4S' is used f or traffic
driven by a Scalable congestion control that also complies with the driven by a Scalable congestion control that also complies with the
additional 'Prague L4S' requirements <xref target="I-D.ietf-tsvwg-ecn-l4 additional 'Prague L4S requirements' <xref target="RFC9331" format="defa
s-id" format="default"/>.</t> ult"/>.</t>
<t>For safe co-existence, under stationary conditions, a Scalable flow <t>For safe coexistence, under stationary conditions, a Scalable flow
has to run at roughly the same rate as a Reno TCP flow (all other has to run at roughly the same rate as a Reno TCP flow (all other
factors being equal). So the drop or marking probability for Classic factors being equal). So the drop or marking probability for Classic
traffic, p_C has to be distinct from the marking probability for L4S traffic, p_C, has to be distinct from the marking probability for L4S
traffic, p_L. The original ECN specification <xref target="RFC3168" form traffic, p_L. The original ECN spec <xref target="RFC3168" format="defau
at="default"/> required these probabilities to be the same, but lt"/> required these probabilities to be the same, but
<xref target="RFC8311" format="default"/> updates RFC 3168 to enable exp <xref target="RFC8311" format="default"/> updates <xref target="RFC3168"
eriments in format="default"/> to enable experiments in
which these probabilities are different.</t> which these probabilities are different.</t>
<t>Also, to remain stable, Classic sources need the network to smooth <t>Also, to remain stable, Classic sources need the network to smooth
p_C so it changes relatively slowly. It is hard for a network node to p_C so it changes relatively slowly. It is hard for a network node to
know the RTTs of all the flows, so a Classic AQM adds a <em>worst-case</ em> RTT of smoothing delay (about 100-200 know the RTTs of all the flows, so a Classic AQM adds a <em>worst-case</ em> RTT of smoothing delay (about 100-200
ms). In contrast, L4S shifts responsibility for smoothing ECN feedback ms). In contrast, L4S shifts responsibility for smoothing ECN feedback
to the sender, which only delays its response by its <em>own</em> RTT, a s well as allowing a more immediate to the sender, which only delays its response by its <em>own</em> RTT, a s well as allowing a more immediate
response if necessary.</t> response if necessary.</t>
<t>The Coupled AQM achieves safe coexistence by making the Classic <t>The Coupled AQM achieves safe coexistence by making the Classic
drop probability p_C proportional to the square of the coupled L4S drop probability p_C proportional to the square of the coupled L4S
probability p_CL. p_CL is an input to the instantaneous L4S marking probability p_CL. p_CL is an input to the instantaneous L4S marking
probability p_L but it changes as slowly as p_C. This makes the Reno probability p_L, but it changes as slowly as p_C. This makes the Reno
flow rate roughly equal the DCTCP flow rate, because the squaring of flow rate roughly equal the DCTCP flow rate, because the squaring of
p_CL counterbalances the square root of p_C in the 'TCP formula' of p_CL counterbalances the square root of p_C in the 'TCP formula' of
Classic Reno congestion control.</t> Classic Reno congestion control.</t>
<t>Stating this as a formula, the relation between Classic drop <t>Stating this as a formula, the relation between Classic drop
probability, p_C, and the coupled L4S probability p_CL needs to take probability, p_C, and the coupled L4S probability p_CL needs to take
the form:</t> the following form:</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ p_C = ( p_CL /
k )^2 (1)]]></artwork> <sourcecode><![CDATA[
p_C = ( p_CL / k )^2, (1)]]></sourcecode>
<t>where k is the constant of proportionality, which is termed the <t>where k is the constant of proportionality, which is termed the
coupling factor.</t> 'coupling factor'.</t>
</section> </section>
<section anchor="dualq" numbered="true" toc="default"> <section anchor="dualq" numbered="true" toc="default">
<name>Dual Queue</name> <name>Dual Queue</name>
<t>Classic traffic needs to build a large queue to prevent <t>Classic traffic needs to build a large queue to prevent
under-utilization. Therefore, a separate queue is provided for L4S underutilization. Therefore, a separate queue is provided for L4S
traffic, and it is scheduled with priority over the Classic queue. traffic, and it is scheduled with priority over the Classic queue.
Priority is conditional to prevent starvation of Classic traffic in Priority is conditional to prevent starvation of Classic traffic in
certain conditions (see <xref target="dualq_coupled_structure" format="d efault"/>).</t> certain conditions (see <xref target="dualq_coupled_structure" format="d efault"/>).</t>
<t>Nonetheless, coupled marking ensures that giving priority to L4S <t>Nonetheless, coupled marking ensures that giving priority to L4S
traffic still leaves the right amount of spare scheduling time for traffic still leaves the right amount of spare scheduling time for
Classic flows to each get equivalent throughput to DCTCP flows (all Classic flows to each get equivalent throughput to DCTCP flows (all
other factors such as RTT being equal).</t> other factors, such as RTT, being equal).</t>
</section> </section>
<section anchor="dualq_classification" numbered="true" toc="default"> <section anchor="dualq_classification" numbered="true" toc="default">
<name>Traffic Classification</name> <name>Traffic Classification</name>
<t>Both the Coupled AQM and DualQ mechanisms need an identifier to <t>Both the Coupled AQM and DualQ mechanisms need an identifier to
distinguish L4S (L) and Classic (C) packets. Then the coupling distinguish L4S (L) and Classic (C) packets.
Then the coupling
algorithm can achieve coexistence without having to inspect flow algorithm can achieve coexistence without having to inspect flow
identifiers, because it can apply the appropriate marking or dropping identifiers, because it can apply the appropriate marking or dropping
probability to all flows of each type. A separate probability to all flows of each type. A separate
specification <xref target="I-D.ietf-tsvwg-ecn-l4s-id" format="default"/ > requires specification <xref target="RFC9331" format="default"/> requires
the network to treat the ECT(1) and CE codepoints of the ECN field as the network to treat the ECT(1) and CE codepoints of the ECN field as
this identifier. An additional process document has proved necessary this identifier. An additional process document has proved necessary
to make the ECT(1) codepoint available for experimentation <xref target= "RFC8311" format="default"/>.</t> to make the ECT(1) codepoint available for experimentation <xref target= "RFC8311" format="default"/>.</t>
<t>For policy reasons, an operator might choose to steer certain <t>For policy reasons, an operator might choose to steer certain
packets (e.g. from certain flows or with certain addresses) out packets (e.g., from certain flows or with certain addresses) out
of the L queue, even though they identify themselves as L4S by their of the L queue, even though they identify themselves as L4S by their
ECN codepoints. In such cases, the L4S ECN protocol <xref target="I-D.ie ECN codepoints. In such cases, the L4S ECN protocol <xref target="RFC933
tf-tsvwg-ecn-l4s-id" format="default"/> says that the device "MUST NOT 1" format="default"/> states that the device "<bcp14>MUST NOT</bcp14>
alter the end-to-end L4S ECN identifier", so that it is preserved alter the end-to-end L4S ECN identifier" so that it is preserved
end-to-end. The aim is that each operator can choose how it treats L4S end to end. The aim is that each operator can choose how it treats L4S
traffic locally, but an individual operator does not alter the traffic locally, but an individual operator does not alter the
identification of L4S packets, which would prevent other operators identification of L4S packets, which would prevent other operators
downstream from making their own choices on how to treat L4S downstream from making their own choices on how to treat L4S
traffic.</t> traffic.</t>
<t>In addition, an operator could use other identifiers to classify <t>In addition, an operator could use other identifiers to classify
certain additional packet types into the L queue that it deems will certain additional packet types into the L queue that it deems will
not risk harm to the L4S service. For instance addresses of specific not risk harm to the L4S service, for instance, addresses of specific
applications or hosts; specific Diffserv codepoints such as EF applications or hosts; specific Diffserv codepoints such as EF, Voice-Ad
(Expedited Forwarding), Voice-Admit or the Non-Queue-Building (NQB) mit, or the Non-Queue-Building (NQB)
per-hop behaviour; or certain protocols (e.g. ARP, DNS) (see per-hop behaviour; or certain protocols (e.g., ARP and DNS) (see <xref t
Section 5.4.1 of <xref target="I-D.ietf-tsvwg-ecn-l4s-id" format="defaul arget="RFC9331" sectionFormat="of" section="5.4.1"/>. Note
t"/>). Note that
that the mechanism only reads these identifiers. <xref target="I-D.ietf- <xref target="RFC9331" format="default"/> states that "a network node <bc
tsvwg-ecn-l4s-id" format="default"/> says it "MUST NOT alter these p14>MUST NOT</bcp14>
non-ECN identifiers". Thus, the L queue is not solely an L4S queue, it change Not-ECT or ECT(0) in the IP-ECN field into an L4S identifier."
can be considered more generally as a low latency queue.</t> Thus, the L queue is not solely an L4S queue; it
can be considered more generally as a low-latency queue.</t>
</section> </section>
<section anchor="dualq_coupled_structure" numbered="true" toc="default"> <section anchor="dualq_coupled_structure" numbered="true" toc="default">
<name>Overall DualQ Coupled AQM Structure</name> <name>Overall DualQ Coupled AQM Structure</name>
<t><xref target="dualq_fig_structure" format="default"/> shows the overa ll structure <t><xref target="dualq_fig_structure" format="default"/> shows the overa ll structure
that any DualQ Coupled AQM is likely to have. This schematic is that any DualQ Coupled AQM is likely to have. This schematic is
intended to aid understanding of the current designs of DualQ Coupled intended to aid understanding of the current designs of DualQ Coupled
AQMs. However, it is not intended to preclude other innovative ways of AQMs. However, it is not intended to preclude other innovative ways of
satisfying the normative requirements in <xref target="dualq_norm_reqs" format="default"/> that minimally define a DualQ Coupled AQM. satisfying the normative requirements in <xref target="dualq_norm_reqs" format="default"/> that minimally define a DualQ Coupled AQM.
Also, the schematic only illustrates operation under normally expected Also, the schematic only illustrates operation under normally expected
circumstances; behaviour under overload or with operator-specific circumstances; behaviour under overload or with operator-specific
classifiers is deferred to <xref target="dualq_unexpected" format="defau lt"/>.</t> classifiers is deferred to <xref target="dualq_unexpected" format="defau lt"/>.</t>
<t>The classifier on the left separates incoming traffic between the <t>The classifier on the left separates incoming traffic between the
two queues (L and C). Each queue has its own AQM that determines the two queues (L and C). Each queue has its own AQM that determines the
likelihood of marking or dropping (p_L and p_C). It has been likelihood of marking or dropping (p_L and p_C).
proved <xref target="PI2" format="default"/> that it is preferable to co In <xref target="PI2" format="default"/>, it has been
ntrol load proved that it is preferable to control load
with a linear controller, then square the output before applying it as with a linear controller, then square the output before applying it as
a drop probability to Reno-friendly traffic (because Reno congestion a drop probability to Reno-friendly traffic (because Reno congestion
control decreases its load proportional to the square-root of the control decreases its load proportional to the square root of the
increase in drop). So, the AQM for Classic traffic needs to be increase in drop). So, the AQM for Classic traffic needs to be
implemented in two stages: i) a base stage that outputs an internal implemented in two stages: i) a base stage that outputs an internal
probability p' (pronounced p-prime); and ii) a squaring stage that probability p' (pronounced 'p-prime') and ii) a squaring stage that
outputs p_C, where</t> outputs p_C, where</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ p_C = (p')^2.
(2)]]></artwork> <sourcecode><![CDATA[
<t>Substituting for p_C in Eqn (1) gives:</t> p_C = (p')^2. (2)]]></sourcecode>
<artwork name="" type="" align="left" alt=""><![CDATA[ p' = p_CL / k] <t>Substituting for p_C in equation (1) gives</t>
]></artwork> <sourcecode><![CDATA[
p' = p_CL / k.]]></sourcecode>
<t>So the slow-moving input to ECN marking in the L queue (the <t>So the slow-moving input to ECN marking in the L queue (the
coupled L4S probability) is:</t> coupled L4S probability) is</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ p_CL = k*p'. <sourcecode><![CDATA[
(3)]]></artwork> p_CL = k*p'. (3)]]></sourcecode>
<t>The actual ECN marking probability p_L that is applied to the L <t>The actual ECN-marking probability p_L that is applied to the L
queue needs to track the immediate L queue delay under L-only queue needs to track the immediate L queue delay under L-only
congestion conditions, as well as track p_CL under coupled congestion congestion conditions, as well as track p_CL under coupled congestion
conditions. So the L queue uses a native AQM that calculates a conditions. So the L queue uses a 'Native AQM' that calculates a
probability p'_L as a function of the instantaneous L queue delay. probability p'_L as a function of the instantaneous L queue delay.
And, given the L queue has conditional priority over the C queue, And given the L queue has conditional priority over the C queue,
whenever the L queue grows, the AQM ought to apply marking probability whenever the L queue grows, the AQM ought to apply marking probability
p'_L, but p_L ought not to fall below p_CL. This suggests:</t> p'_L, but p_L ought to not fall below p_CL. This suggests</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ p_L = max(p'_L <sourcecode><![CDATA[
, p_CL), (4)]]></artwork> p_L = max(p'_L, p_CL), (4)]]></sourcecode>
<t>which has also been found to work very well in <t>which has also been found to work very well in
practice.</t> practice.</t>
<t>The two transformations of p' in equations (2) and (3) implement <t>The two transformations of p' in equations (2) and (3) implement
the required coupling given in equation (1) earlier.</t> the required coupling given in equation (1) earlier.</t>
<t>The constant of proportionality or coupling factor, k, in equation <t>The constant of proportionality or coupling factor, k, in equation
(1) determines the ratio between the congestion probabilities (loss or (1) determines the ratio between the congestion probabilities (loss or
marking) experienced by L4S and Classic traffic. Thus, k indirectly marking) experienced by L4S and Classic traffic. Thus, k indirectly
determines the ratio between L4S and Classic flow rates, because flows determines the ratio between L4S and Classic flow rates, because flows
(assuming they are responsive) adjust their rate in response to (assuming they are responsive) adjust their rate in response to
congestion probability. <xref target="dualq_Choosing_k" format="default" /> gives congestion probability. <xref target="dualq_Choosing_k" format="default" /> gives
skipping to change at line 680 skipping to change at line 645
==>|Classifier| ,-------. (k*p') [ priority]==> ==>|Classifier| ,-------. (k*p') [ priority]==>
| |\ | Base | | \scheduler/ | |\ | Base | | \scheduler/
`----------'\\ | AQM |---->: ,'|`-.___.-' `----------'\\ | AQM |---->: ,'|`-.___.-'
\\ | |p' | <' | \\ | |p' | <' |
\\ `-------' (p'^2) //`' \\ `-------' (p'^2) //`'
\\ ^ | // \\ ^ | //
\\,. | v p_C // \\,. | v p_C //
< | _________ .------.// < | _________ .------.//
`\| | | | Drop |/ `\| | | | Drop |/
Classic (C) |queue |===>|/mark | Classic (C) |queue |===>|/mark |
__|______| `------' __|______| `------']]>
]]></artwork> Legend: ===> traffic flow
---> control dependency
</artwork>
</figure> </figure>
<t keepWithPrevious="true">Legend: ===&gt; traffic flow; ---&gt; control
dependency.</t>
<t>After the AQMs have applied their dropping or marking, the <t>After the AQMs have applied their dropping or marking, the
scheduler forwards their packets to the link. Even though the scheduler forwards their packets to the link. Even though the
scheduler gives priority to the L queue, it is not as strong as the scheduler gives priority to the L queue, it is not as strong as the
coupling from the C queue. This is because, as the C queue grows, the coupling from the C queue. This is because, as the C queue grows, the
base AQM applies more congestion signals to L traffic (as well as C). 'Base AQM' applies more congestion signals to L traffic (as well as to C ).
As L flows reduce their rate in response, they use less than the As L flows reduce their rate in response, they use less than the
scheduling share for L traffic. So, because the scheduler is work scheduling share for L traffic. So, because the scheduler is work
preserving, it schedules any C traffic in the gaps.</t> preserving, it schedules any C traffic in the gaps.</t>
<t>Giving priority to the L queue has the benefit of very low L queue <t>Giving priority to the L queue has the benefit of very low L queue
delay, because the L queue is kept empty whenever L traffic is delay, because the L queue is kept empty whenever L traffic is
controlled by the coupling. Also, there only has to be a coupling in controlled by the coupling. Also, there only has to be a coupling in
one direction - from Classic to L4S. Priority has to be conditional in one direction -- from Classic to L4S. Priority has to be conditional in
some way to prevent the C queue being starved in the short-term (see some way to prevent the C queue from being starved in the short term (se
e
<xref target="dualq_Overload_Starvation" format="default"/>) to give C t raffic a means <xref target="dualq_Overload_Starvation" format="default"/>) to give C t raffic a means
to push in, as explained next. With normal responsive L traffic, the to push in, as explained next. With normal responsive L traffic, the
coupled ECN marking gives C traffic the ability to push back against coupled ECN marking gives C traffic the ability to push back against
even strict priority, by congestion marking the L traffic to make it even strict priority, by congestion marking the L traffic to make it
yield some space. However, if there is just a small finite set of C yield some space. However, if there is just a small finite set of C
packets (e.g. a DNS request or an initial window of data) some packets (e.g., a DNS request or an initial window of data), some
Classic AQMs will not induce enough ECN marking in the L queue, no Classic AQMs will not induce enough ECN marking in the L queue, no
matter how long the small set of C packets waits. Then, if the L queue matter how long the small set of C packets waits. Then, if the L queue
happens to remain busy, the C traffic would never get a scheduling happens to remain busy, the C traffic would never get a scheduling
opportunity from a strict priority scheduler. Ideally the Classic AQM opportunity from a strict priority scheduler. Ideally, the Classic AQM
would be designed to increase the coupled marking the longer that C would be designed to increase the coupled marking the longer that C
packets have been waiting, but this is not always practical - hence packets have been waiting, but this is not always practical -- hence
the need for L priority to be conditional. Giving a small weight or the need for L priority to be conditional. Giving a small weight or
limited waiting time for C traffic improves response times for short limited waiting time for C traffic improves response times for short
Classic messages, such as DNS requests, and improves Classic flow Classic messages, such as DNS requests, and improves Classic flow
startup because immediate capacity is available.</t> startup because immediate capacity is available.</t>
<t>Example DualQ Coupled AQM algorithms called DualPI2 and Curvy RED <t>Example DualQ Coupled AQM algorithms called 'DualPI2' and 'Curvy RED'
are given in <xref target="dualq_Ex_algo_pi2" format="default"/> and <xr are given in Appendices <xref target="dualq_Ex_algo_pi2" format="counter
ef target="dualq_Ex_algo" format="default"/>. Either example AQM can be used to "/> and <xref target="dualq_Ex_algo" format="counter"/>. Either example AQM can
couple be used to couple
packet marking and dropping across a dual Q.</t> packet marking and dropping across a DualQ:</t>
<t>DualPI2 uses a Proportional-Integral (PI) controller as the Base <ul spacing="normal">
<li><t>DualPI2 uses a Proportional Integral (PI) controller as the Base
AQM. Indeed, this Base AQM with just the squared output and no L4S AQM. Indeed, this Base AQM with just the squared output and no L4S
queue can be used as a drop-in replacement for PIE <xref target="RFC8033 queue can be used as a drop-in replacement for PIE <xref target="RFC8033
" format="default"/>, in which case it is just called PI2 <xref target="PI2" for " format="default"/>, in which case it is just called PI2 <xref target="PI2" for
mat="default"/>. PI2 is a principled simplification of PIE that is both mat="default"/>.
PI2 is a principled simplification of PIE that is both
more responsive and more stable in the face of dynamically varying more responsive and more stable in the face of dynamically varying
load.</t> load.</t></li>
<t>Curvy RED is derived from RED <xref target="RFC2309" format="default" <li><t>Curvy RED is derived from RED <xref target="RED" format="default"
/>, except />, except
its configuration parameters are delay-based to make them insensitive its configuration parameters are delay-based to make them insensitive
to link rate and it requires fewer operations per packet than RED. to link rate, and it requires fewer operations per packet than RED.
However, DualPI2 is more responsive and stable over a wider range of However, DualPI2 is more responsive and stable over a wider range of
RTTs than Curvy RED. As a consequence, at the time of writing, DualPI2 RTTs than Curvy RED. As a consequence, at the time of writing, DualPI2
has attracted more development and evaluation attention than Curvy has attracted more development and evaluation attention than Curvy
RED, leaving the Curvy RED design not so fully evaluated.</t> RED, leaving the Curvy RED design not so fully evaluated.</t></li>
</ul>
<t>Both AQMs regulate their queue against targets configured in units <t>Both AQMs regulate their queue against targets configured in units
of time rather than bytes. As already explained, this ensures of time rather than bytes. As already explained, this ensures
configuration can be invariant for different drain rates. With AQMs in configuration can be invariant for different drain rates. With AQMs in
a dualQ structure this is particularly important because the drain a DualQ structure this is particularly important because the drain
rate of each queue can vary rapidly as flows for the two queues arrive rate of each queue can vary rapidly as flows for the two queues arrive
and depart, even if the combined link rate is constant.</t> and depart, even if the combined link rate is constant.</t>
<t>It would be possible to control the queues with other alternative <t>It would be possible to control the queues with other alternative
AQMs, as long as the normative requirements (those expressed in AQMs, as long as the normative requirements (those expressed in
capitals) in <xref target="dualq_norm_reqs" format="default"/> are obser ved.</t> capitals) in <xref target="dualq_norm_reqs" format="default"/> are obser ved.</t>
<t>The two queues could optionally be part of a larger queuing <t>The two queues could optionally be part of a larger queuing
hierarchy, such as the initial example ideas in <xref target="I-D.brisco e-tsvwg-l4s-diffserv" format="default"/>.</t> hierarchy, such as the initial example ideas in <xref target="I-D.brisco e-tsvwg-l4s-diffserv" format="default"/>.</t>
</section> </section>
<section anchor="dualq_norm_reqs" numbered="true" toc="default"> <section anchor="dualq_norm_reqs" numbered="true" toc="default">
<name>Normative Requirements for a DualQ Coupled AQM</name> <name>Normative Requirements for a DualQ Coupled AQM</name>
<t>The following requirements are intended to capture only the <t>The following requirements are intended to capture only the
essential aspects of a DualQ Coupled AQM. They are intended to be essential aspects of a DualQ Coupled AQM. They are intended to be
independent of the particular AQMs implemented for each queue, but to independent of the particular AQMs implemented for each queue but to
still define the DualQ framework built around those AQMs.</t> still define the DualQ framework built around those AQMs.</t>
<section anchor="dualq_functional_reqs" numbered="true" toc="default"> <section anchor="dualq_functional_reqs" numbered="true" toc="default">
<name>Functional Requirements</name> <name>Functional Requirements</name>
<t>A Dual Queue Coupled AQM implementation MUST comply with the <t>A DualQ Coupled AQM implementation <bcp14>MUST</bcp14> comply with the
prerequisite L4S behaviours for any L4S network node (not just a prerequisite L4S behaviours for any L4S network node (not just a
DualQ) as specified in section 5 of <xref target="I-D.ietf-tsvwg-ecn-l DualQ) as specified in <xref
4s-id" format="default"/>. These primarily concern target="RFC9331" sectionFormat="of" section="5"/>. These primarily concern
classification and remarking as briefly summarized in <xref target="du classification and re-marking as briefly summarized earlier in <xref t
alq_classification" format="default"/> earlier. But there is also a arget="dualq_classification" format="default"/>. But
subsection (5.5) giving guidance on reducing the burstiness of the <xref target="RFC9331" sectionFormat="of" section="5.5"/> also gives
guidance on reducing the burstiness of the
link technology underlying any L4S AQM.</t> link technology underlying any L4S AQM.</t>
<t>A Dual Queue Coupled AQM implementation MUST utilize two queues, <t>A DualQ Coupled AQM implementation <bcp14>MUST</bcp14> utilize two queues,
each with an AQM algorithm.</t> each with an AQM algorithm.</t>
<t>The AQM algorithm for the low latency (L) queue MUST be able to <t>The AQM algorithm for the low-latency (L) queue <bcp14>MUST</bcp14> be able to
apply ECN marking to ECN-capable packets.</t> apply ECN marking to ECN-capable packets.</t>
<t>The scheduler draining the two queues MUST give L4S packets <t>The scheduler draining the two queues <bcp14>MUST</bcp14> give L4S
priority over Classic, although priority MUST be bounded in order packets
not to starve Classic traffic (see <xref target="dualq_Overload_Starva priority over Classic, although priority <bcp14>MUST</bcp14> be bounde
tion" format="default"/>). The scheduler SHOULD be d in order
not to starve Classic traffic (see <xref target="dualq_Overload_Starva
tion" format="default"/>). The scheduler <bcp14>SHOULD</bcp14> be
work-conserving, or otherwise close to work-conserving. This is work-conserving, or otherwise close to work-conserving. This is
because Classic traffic needs to be able to efficiently fill any because Classic traffic needs to be able to efficiently fill any
space left by L4S traffic even though the scheduler would otherwise space left by L4S traffic even though the scheduler would otherwise
allocate it to L4S.</t> allocate it to L4S.</t>
<t><xref target="I-D.ietf-tsvwg-ecn-l4s-id" format="default"/> defines the meaning of <t><xref target="RFC9331" format="default"/> defines the meaning of
an ECN marking on L4S traffic, relative to drop of Classic traffic. an ECN marking on L4S traffic, relative to drop of Classic traffic.
In order to ensure coexistence of Classic and Scalable L4S traffic, In order to ensure coexistence of Classic and Scalable L4S traffic,
it says, "The likelihood that an AQM drops a Not-ECT Classic packet it says,
(p_C) MUST be roughly proportional to the square of the likelihood "the likelihood that the AQM drops a Not-ECT Classic packet
(p_C) <bcp14>MUST</bcp14> be roughly proportional to the square of the
likelihood
that it would have marked it if it had been an L4S packet (p_L)." that it would have marked it if it had been an L4S packet (p_L)."
The term 'likelihood' is used to allow for marking and dropping to The term 'likelihood' is used to allow for marking and dropping to
be either probabilistic or deterministic.</t> be either probabilistic or deterministic.</t>
<t>For the current specification, this translates into the following <t>For the current specification, this translates into the following
requirement. A DualQ Coupled AQM MUST apply ECN marking to traffic requirement. A DualQ Coupled AQM <bcp14>MUST</bcp14> apply ECN marking to traffic
in the L queue that is no lower than that derived from the in the L queue that is no lower than that derived from the
likelihood of drop (or ECN marking) in the Classic queue using Eqn. likelihood of drop (or ECN marking) in the Classic queue using equatio n
(1).</t> (1).</t>
<t>The constant of proportionality, k, in Eqn (1) determines the <t>The constant of proportionality, k, in equation (1) determines the
relative flow rates of Classic and L4S flows when the AQM concerned relative flow rates of Classic and L4S flows when the AQM concerned
is the bottleneck (all other factors being equal). The L4S ECN is the bottleneck (all other factors being equal). The L4S ECN
protocol <xref target="I-D.ietf-tsvwg-ecn-l4s-id" format="default"/> s protocol <xref target="RFC9331" format="default"/> says,
ays, "The
"The
constant of proportionality (k) does not have to be standardised for constant of proportionality (k) does not have to be standardised for
interoperability, but a value of 2 is RECOMMENDED."</t> interoperability, but a value of 2 is <bcp14>RECOMMENDED</bcp14>."
</t>
<t>Assuming Scalable congestion controls for the Internet will be as <t>Assuming Scalable congestion controls for the Internet will be as
aggressive as DCTCP, this will ensure their congestion window will aggressive as DCTCP, this will ensure their congestion window will
be roughly the same as that of a standards track TCP Reno congestion be roughly the same as that of a Standards Track TCP Reno congestion
control (Reno) <xref target="RFC5681" format="default"/> and other Ren control (Reno) <xref target="RFC5681" format="default"/> and other Ren
o-friendly o-friendly
controls, such as TCP Cubic in its Reno-compatibility mode.</t> controls, such as TCP CUBIC in its Reno-friendly mode.</t>
<!--{ToDo: The TCP Prague requirements are not necessarily final.
If the aggressiveness of DCTCP is not defined as the benchmark for Scalable cont
rols on
the Internet, the recommended value of k will also be subject to change.}-->
<t>The choice of k is a matter of operator policy, and operators MAY <t>The choice of k is a matter of operator policy, and operators <bcp1 4>MAY</bcp14>
choose a different value using the guidelines in <xref target="dualq_C hoosing_k" format="default"/>.</t> choose a different value using the guidelines in <xref target="dualq_C hoosing_k" format="default"/>.</t>
<t>If multiple customers or users share capacity at a bottleneck <t>If multiple customers or users share capacity at a bottleneck
(e.g. in the Internet access link of a campus network), the (e.g., in the Internet access link of a campus network), the
operator's choice of k will determine capacity sharing between the operator's choice of k will determine capacity sharing between the
flows of different customers. However, on the public Internet, flows of different customers. However, on the public Internet,
access network operators typically isolate customers from each other access network operators typically isolate customers from each other
with some form of layer-2 multiplexing (OFDM(A) in DOCSIS3.1, CDMA with some form of Layer 2 multiplexing
in 3G, SC-FDMA in LTE) or L3 scheduling (WRR in DSL), rather than (OFDM(A) in DOCSIS 3.1,
CDMA in 3G, and SC-FDMA in LTE) or Layer 3 scheduling (Weighted Round Robin (WR
R) for DSL) rather than
relying on host congestion controls to share capacity between relying on host congestion controls to share capacity between
customers <xref target="RFC0970" format="default"/>. In such cases, th e choice customers <xref target="RFC0970" format="default"/>. In such cases, th e choice
of k will solely affect relative flow rates within each customer's of k will solely affect relative flow rates within each customer's
access capacity, not between customers. Also, k will not affect access capacity, not between customers. Also, k will not affect
relative flow rates at any times when all flows are Classic or all relative flow rates at any times when all flows are Classic or all
flows are L4S, and it will not affect the relative throughput of flows are L4S, and it will not affect the relative throughput of
small flows.</t> small flows.</t>
<t/> <t/>
<section anchor="dualq_unexpected" numbered="true" toc="default"> <section anchor="dualq_unexpected" numbered="true" toc="default">
<name>Requirements in Unexpected Cases</name> <name>Requirements in Unexpected Cases</name>
<t>The flexibility to allow operator-specific classifiers (<xref tar get="dualq_classification" format="default"/>) leads to the need to specify what <t>The flexibility to allow operator-specific classifiers (<xref tar get="dualq_classification" format="default"/>) leads to the need to specify what
the AQM in each queue ought to do with packets that do not carry the AQM in each queue ought to do with packets that do not carry
the ECN field expected for that queue. It is expected that the AQM the ECN field expected for that queue. It is expected that the AQM
in each queue will inspect the ECN field to determine what sort of in each queue will inspect the ECN field to determine what sort of
congestion notification to signal, then it will decide whether to congestion notification to signal, then it will decide whether to
apply congestion notification to this particular packet, as apply congestion notification to this particular packet, as
follows:</t> follows:</t>
<ul spacing="normal"> <ul spacing="normal">
<li> <li>
<t>If a packet that does not carry an ECT(1) or CE codepoint <t>If a packet that does not carry an ECT(1) or a CE codepoint
is classified into the L queue:</t> is classified into the L queue, then:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>if the packet is ECT(0), the L AQM SHOULD apply <li>if the packet is ECT(0), the L AQM <bcp14>SHOULD</bcp14> a
CE-marking using a probability appropriate to Classic pply
CE marking using a probability appropriate to Classic
congestion control and appropriate to the target delay in congestion control and appropriate to the target delay in
the L queue</li> the L queue</li>
<li> <li>
<t>if the packet is Not-ECT, the appropriate action <t>if the packet is Not-ECT, the appropriate action
depends on whether some other function is protecting the L depends on whether some other function is protecting the L
queue from misbehaving flows (e.g. per-flow queue queue from misbehaving flows (e.g., per-flow queue
protection <xref target="I-D.briscoe-docsis-q-protection" fo rmat="default"/> or latency protection <xref target="I-D.briscoe-docsis-q-protection" fo rmat="default"/> or latency
policing):</t> policing):</t>
<ul spacing="normal"> <ul spacing="normal">
<li>If separate queue protection is provided, the L AQM <li>if separate queue protection is provided, the L AQM
SHOULD ignore the packet and forward it unchanged, <bcp14>SHOULD</bcp14> ignore the packet and forward it u
nchanged,
meaning it should not calculate whether to apply meaning it should not calculate whether to apply
congestion notification and it should neither drop nor congestion notification, and it should neither drop nor
CE-mark the packet (for instance, the operator might CE mark the packet (for instance, the operator might
classify EF traffic that is unresponsive to drop into classify EF traffic that is unresponsive to drop into
the L queue, alongside responsive L4S-ECN traffic)</li> the L queue, alongside responsive L4S-ECN traffic)</li>
<li>if separate queue protection is not provided, the L <li>if separate queue protection is not provided, the L
AQM SHOULD apply drop using a drop probability AQM <bcp14>SHOULD</bcp14> apply drop using a drop probab ility
appropriate to Classic congestion control and appropriate to Classic congestion control and
appropriate to the target delay in the L queue</li> to the target delay in the L queue</li>
</ul> </ul>
</li> </li>
</ul> </ul>
</li> </li>
<li> <li>
<t>If a packet that carries an ECT(1) codepoint is classified <t>If a packet that carries an ECT(1) codepoint is classified
into the C queue:</t> into the C queue:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>the C AQM SHOULD apply CE-marking using the coupled AQM <li>the C AQM <bcp14>SHOULD</bcp14> apply CE marking using the Coupled AQM
probability p_CL (= k*p').</li> probability p_CL (= k*p').</li>
</ul> </ul>
</li> </li>
</ul> </ul>
<t>The above requirements are worded as "SHOULDs", because <t>The above requirements are worded as "<bcp14>SHOULD</bcp14>"s, be cause
operator-specific classifiers are for flexibility, by definition. operator-specific classifiers are for flexibility, by definition.
Therefore, alternative actions might be appropriate in the Therefore, alternative actions might be appropriate in the
operator's specific circumstances. An example would be where the operator's specific circumstances.
operator knows that certain legacy traffic marked with one An example would be where the
operator knows that certain legacy traffic set to one
codepoint actually has a congestion response associated with codepoint actually has a congestion response associated with
another codepoint.</t> another codepoint.</t>
<t>If the DualQ Coupled AQM has detected overload, it MUST <t>If the DualQ Coupled AQM has detected overload, it <bcp14>MUST</b cp14>
introduce Classic drop to both types of ECN-capable traffic until introduce Classic drop to both types of ECN-capable traffic until
the overload episode has subsided. Introducing drop if ECN marking the overload episode has subsided. Introducing drop if ECN marking
is persistently high is recommended by Section 7 of the ECN is persistently high is recommended in
specification <xref target="RFC3168" format="default"/> and Section
4.2.1 of Section <xref target="RFC3168" sectionFormat="bare" section="7"/> of
the AQM Recommendations <xref target="RFC7567" format="default"/>.</ the ECN spec <xref target="RFC3168"/>
t> and in Section <xref target="RFC7567" sectionFormat="bare" section="
4.2.1"/> of
the AQM Recommendations <xref target="RFC7567"/>.</t>
</section> </section>
</section> </section>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Management Requirements</name> <name>Management Requirements</name>
<t/> <t/>
<section anchor="dualq_config" numbered="true" toc="default"> <section anchor="dualq_config" numbered="true" toc="default">
<name>Configuration</name> <name>Configuration</name>
<t>By default, a DualQ Coupled AQM SHOULD NOT need any <t>By default, a DualQ Coupled AQM <bcp14>SHOULD NOT</bcp14> need an y
configuration for use at a bottleneck on the public configuration for use at a bottleneck on the public
Internet <xref target="RFC7567" format="default"/>. The following pa Internet <xref target="RFC7567" format="default"/>. The following pa
rameters rameters
MAY be operator-configurable, e.g. to tune for non-Internet <bcp14>MAY</bcp14> be operator-configurable, e.g., to tune for non-I
nternet
settings:</t> settings:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>Optional packet classifier(s) to use in addition to the ECN <li>Optional packet classifier(s) to use in addition to the ECN
field (see <xref target="dualq_classification" format="default"/ >);</li> field (see <xref target="dualq_classification" format="default"/ >).</li>
<li> <li>
<t>Expected typical RTT, which can be used to determine the <t>Expected typical RTT, which can be used to determine the
queuing delay of the Classic AQM at its operating point, in queuing delay of the Classic AQM at its operating point, in
order to prevent typical lone flows from under-utilizing order to prevent typical lone flows from underutilizing
capacity. For example:</t> capacity. For example:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>for the PI2 algorithm (<xref target="dualq_Ex_algo_pi2" fo <li>for the PI2 algorithm (<xref target="dualq_Ex_algo_pi2" fo
rmat="default"/>) the queuing delay target is rmat="default"/>), the queuing delay target is
dependent on the typical RTT;</li> dependent on the typical RTT.</li>
<li>for the Curvy RED algorithm (<xref target="dualq_Ex_algo" <li>for the Curvy RED algorithm (<xref target="dualq_Ex_algo"
format="default"/>) the queuing delay at the desired format="default"/>), the queuing delay at the desired
operating point of the curvy ramp is configured to operating point of the curvy ramp is configured to
encompass a typical RTT;</li> encompass a typical RTT.</li>
<li>if another Classic AQM was used, it would be likely to <li>if another Classic AQM was used, it would be likely to
need an operating point for the queue based on the typical need an operating point for the queue based on the typical
RTT, and if so it SHOULD be expressed in units of RTT, and if so, it <bcp14>SHOULD</bcp14> be expressed in uni ts of
time.</li> time.</li>
</ul> </ul>
<t>An operating point that is manually calculated might <t>An operating point that is manually calculated might
be directly configurable instead, e.g. for links with be directly configurable instead, e.g., for links with
large numbers of flows where under-utilization by a single large numbers of flows where underutilization by a single
flow would be unlikely.</t> flow would be unlikely.</t>
</li> </li>
<li> <li>
<t>Expected maximum RTT, which can be used to set the <t>Expected maximum RTT, which can be used to set the
stability parameter(s) of the Classic AQM. For example:</t> stability parameter(s) of the Classic AQM. For example:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>for the PI2 algorithm (<xref target="dualq_Ex_algo_pi2" fo rmat="default"/>), the gain parameters of the <li>for the PI2 algorithm (<xref target="dualq_Ex_algo_pi2" fo rmat="default"/>), the gain parameters of the
PI algorithm depend on the maximum RTT.</li> PI algorithm depend on the maximum RTT.</li>
<li>for the Curvy RED algorithm (<xref target="dualq_Ex_algo" format="default"/>) the smoothing parameter is <li>for the Curvy RED algorithm (<xref target="dualq_Ex_algo" format="default"/>), the smoothing parameter is
chosen to filter out transients in the queue within a chosen to filter out transients in the queue within a
maximum RTT.</li> maximum RTT.</li>
</ul> </ul>
<t>Stability parameter(s) that are manually calculated <t>Any stability parameter that is manually calculated
assuming a maximum RTT might be directly configurable assuming a maximum RTT might be directly configurable
instead.</t> instead.</t>
</li> </li>
<li>Coupling factor, k (see <xref target="dualq_Choosing_k" format ="default"/>);</li> <li>Coupling factor, k (see <xref target="dualq_Choosing_k" format ="default"/>).</li>
<li> <li>
<t>A limit to the conditional priority of L4S. This is <t>A limit to the conditional priority of L4S. This is
scheduler-dependent, but it SHOULD be expressed as a relation scheduler-dependent, but it <bcp14>SHOULD</bcp14> be expressed a s a relation
between the max delay of a C packet and an L packet. For between the max delay of a C packet and an L packet. For
example:</t> example:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>for a WRR scheduler a weight ratio between L and C of <li>for a WRR scheduler, a weight ratio between L and C of
w:1 means that the maximum delay to a C packet is w times w:1 means that the maximum delay of a C packet is w times
that of an L packet.</li> that of an L packet.</li>
<li>for a time-shifted FIFO (TS-FIFO) scheduler (see <xref tar get="dualq_Overload_Starvation" format="default"/>) a time-shift of <li>for a time-shifted FIFO (TS-FIFO) scheduler (see <xref tar get="dualq_Overload_Starvation" format="default"/>), a time-shift of
tshift means that the maximum delay to a C packet is tshift means that the maximum delay to a C packet is
tshift greater than that of an L packet. tshift could be tshift greater than that of an L packet. tshift could be
expressed as a multiple of the typical RTT rather than as expressed as a multiple of the typical RTT rather than as
an absolute delay.</li> an absolute delay.</li>
</ul> </ul>
</li> </li>
<li>The maximum Classic ECN marking probability, p_Cmax, before <li>The maximum Classic ECN-marking probability, p_Cmax, before
introducing drop.</li> introducing drop.</li>
</ul> </ul>
</section> </section>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Monitoring</name> <name>Monitoring</name>
<t>An experimental DualQ Coupled AQM SHOULD allow the operator to <t>An experimental DualQ Coupled AQM <bcp14>SHOULD</bcp14> allow the operator to
monitor each of the following operational statistics on demand, monitor each of the following operational statistics on demand,
per queue and per configurable sample interval, for performance per queue and per configurable sample interval, for performance
monitoring and perhaps also for accounting in some cases:</t> monitoring and perhaps also for accounting in some cases:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>Bits forwarded, from which utilization can be <li>bits forwarded, from which utilization can be
calculated;</li> calculated;</li>
<li>Total packets in the three categories: arrived, presented <li>total packets in the three categories: arrived, presented
to the AQM, and forwarded. The difference between the first to the AQM, and forwarded. The difference between the first
two will measure any non-AQM tail discard. The difference two will measure any non-AQM tail discard. The difference
between the last two will measure proactive AQM discard;</li> between the last two will measure proactive AQM discard;</li>
<li>ECN packets marked, non-ECN packets dropped, ECN packets <li>ECN packets marked, non-ECN packets dropped, and ECN packets
dropped, which can be combined with the three total packet dropped, which can be combined with the three total packet
counts above to calculate marking and dropping counts above to calculate marking and dropping
probabilities;</li> probabilities; and</li>
<li> <li>
<t>Queue delay (not including serialization delay of the head <t>queue delay (not including serialization delay of the head
packet or medium acquisition delay) - see further notes packet or medium acquisition delay) -- see further notes
below.</t> below.</t>
<t>Unlike the other statistics, <t>Unlike the other statistics,
queue delay cannot be captured in a simple accumulating queue delay cannot be captured in a simple accumulating
counter. Therefore, the type of queue delay statistics counter. Therefore, the type of queue delay statistics
produced (mean, percentiles, etc.) will depend on produced (mean, percentiles, etc.) will depend on
implementation constraints. To facilitate comparative implementation constraints. To facilitate comparative
evaluation of different implementations and approaches, an evaluation of different implementations and approaches, an
implementation SHOULD allow mean and 99th percentile queue implementation <bcp14>SHOULD</bcp14> allow mean and 99th percent ile queue
delay to be derived (per queue per sample interval). A delay to be derived (per queue per sample interval). A
relatively simple way to do this would be to store a relatively simple way to do this would be to store a
coarse-grained histogram of queue delay. This could be done coarse-grained histogram of queue delay. This could be done
with a small number of bins with configurable edges that with a small number of bins with configurable edges that
represent contiguous ranges of queue delay. Then, over a represent contiguous ranges of queue delay. Then, over a
sample interval, each bin would accumulate a count of the sample interval, each bin would accumulate a count of the
number of packets that had fallen within each range. The number of packets that had fallen within each range. The
maximum queue delay per queue per interval MAY also be maximum queue delay per queue per interval <bcp14>MAY</bcp14> al so be
recorded, to aid diagnosis of faults and anomalous events.</t> recorded, to aid diagnosis of faults and anomalous events.</t>
</li> </li>
</ul> </ul>
</section> </section>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Anomaly Detection</name> <name>Anomaly Detection</name>
<t>An experimental DualQ Coupled AQM SHOULD asynchronously report <t>An experimental DualQ Coupled AQM <bcp14>SHOULD</bcp14> asynchron ously report
the following data about anomalous conditions:</t> the following data about anomalous conditions:</t>
<ul spacing="normal"> <ul spacing="normal">
<li> <li>
<t>Start-time and duration of overload state.</t> <t>Start time and duration of overload state.</t>
<t>A hysteresis mechanism SHOULD be used to <t>A hysteresis mechanism <bcp14>SHOULD</bcp14> be used to
prevent flapping in and out of overload causing an event prevent flapping in and out of overload causing an event
storm. For instance, exit from overload state could trigger storm. For instance, exiting from overload state could trigger
one report, but also latch a timer. Then, during that time, if one report but also latch a timer. Then, during that time, if
the AQM enters and exits overload state any number of times, the AQM enters and exits overload state any number of times,
the duration in overload state is accumulated, but no new the duration in overload state is accumulated, but no new
report is generated until the first time the AQM is out of report is generated until the first time the AQM is out of
overload once the timer has expired.</t> overload once the timer has expired.</t>
</li> </li>
</ul> </ul>
</section> </section>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Deployment, Coexistence and Scaling</name> <name>Deployment, Coexistence, and Scaling</name>
<t><xref target="RFC5706" format="default"/> suggests that deploymen <t><xref target="RFC5706" format="default"/> suggests that deploymen
t, coexistence t, coexistence,
and scaling should also be covered as management requirements. The and scaling should also be covered as management requirements. The
raison d'etre of the DualQ Coupled AQM is to enable raison d'etre of the DualQ Coupled AQM is to enable
deployment and coexistence of Scalable congestion controls - as deployment and coexistence of Scalable congestion controls (as
incremental replacements for today's Reno-friendly controls that incremental replacements for today's Reno-friendly controls that
do not scale with bandwidth-delay product. Therefore, there is no do not scale with bandwidth-delay product). Therefore, there is no
need to repeat these motivating issues here given they are already need to repeat these motivating issues here given they are already
explained in the Introduction and detailed in the L4S explained in the Introduction and detailed in the L4S
architecture <xref target="I-D.ietf-tsvwg-l4s-arch" format="default" />.</t> architecture <xref target="RFC9330" format="default"/>.</t>
<t>The descriptions of specific DualQ Coupled AQM algorithms in <t>The descriptions of specific DualQ Coupled AQM algorithms in
the appendices cover scaling of their configuration parameters, the appendices cover scaling of their configuration parameters,
e.g. with respect to RTT and sampling frequency.</t> e.g., with respect to RTT and sampling frequency.</t>
</section> </section>
</section> </section>
</section> </section>
</section> </section>
<section anchor="dualq_IANA" numbered="true" toc="default"> <section anchor="dualq_IANA" numbered="true" toc="default">
<name>IANA Considerations (to be removed by RFC Editor)</name> <name>IANA Considerations</name>
<t>This specification contains no IANA considerations.</t> <t>This document has no IANA actions.</t>
</section> </section>
<section anchor="dualq_Security_Considerations" numbered="true" toc="default "> <section anchor="dualq_Security_Considerations" numbered="true" toc="default ">
<name>Security Considerations</name> <name>Security Considerations</name>
<t/> <t/>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Low Delay without Requiring Per-Flow Processing</name> <name>Low Delay without Requiring Per-flow Processing</name>
<t>The L4S architecture <xref target="I-D.ietf-tsvwg-l4s-arch" format="d <t>The L4S architecture <xref target="RFC9330" format="default"/>
efault"/> compares the DualQ and FQ approaches to L4S. The
compares the DualQ and per-flow-queuing (FQ) approaches to L4S. The
privacy considerations section in that document motivates the DualQ on privacy considerations section in that document motivates the DualQ on
the grounds that users who want to encrypt application flow the grounds that users who want to encrypt application flow
identifiers, e.g. in IPSec or other encrypted VPN tunnels, don't identifiers, e.g., in IPsec or other encrypted VPN tunnels, don't
have to sacrifice low delay (<xref target="RFC8404" format="default"/> e ncourages have to sacrifice low delay (<xref target="RFC8404" format="default"/> e ncourages
avoidance of such privacy compromises).</t> avoidance of such privacy compromises).</t>
<t>The security considerations section of the L4S architecture also <t>The security considerations section of the L4S architecture <xref tar
includes subsections on policing of relative flow-rates (section 8.1) get="RFC9330" format="default"/> also
and on policing of flows that cause excessive queuing delay (section includes subsections on policing of relative flow rates (Section <xref
8.2). It explains that the interests of users do not collide in the target="RFC9330" sectionFormat="bare" section="8.1"/>) and on
same way for delay as they do for bandwidth. For someone to get more policing of flows that cause excessive queuing delay (Section <xref
of the bandwidth of a shared link, someone else necessarily gets less target="RFC9330" sectionFormat="bare" section="8.2"/>). It explains
(a 'zero-sum game'), whereas queuing delay can be reduced for that the interests of users do not collide in the same way for delay
everyone, without any need for someone else to lose out. It also as they do for bandwidth. For someone to get more of the bandwidth of
explains that, on the current Internet, scheduling usually enforces a shared link, someone else necessarily gets less (a 'zero-sum game'),
separation of bandwidth between 'sites' (e.g. households, whereas queuing delay can be reduced for everyone, without any need
businesses or mobile users), but it is not common to need to schedule for someone else to lose out. It also explains that, on the current
or police the bandwidth used by individual application flows.</t> Internet, scheduling usually enforces separation of bandwidth between
'sites' (e.g., households, businesses, or mobile users), but it is not
common to need to schedule or police the bandwidth used by individual
application flows.</t>
<t>By the above arguments, per-flow rate policing might not be <t>By the above arguments, per-flow rate policing might not be
necessary and in trusted environments (e.g. private data centres) necessary, and in trusted environments (e.g., private data centres),
it is certainly unlikely to be needed. Therefore, because it is hard it is certainly unlikely to be needed. Therefore, because it is hard
to avoid complexity and unintended side effects with per-flow rate to avoid complexity and unintended side effects with per-flow rate
policing, it needs to be separable from a basic AQM, as an option, policing, it needs to be separable from a basic AQM, as an option,
under policy control. On this basis, the DualQ Coupled AQM provides under policy control. On this basis, the DualQ Coupled AQM provides
low delay without prejudging the question of per-flow rate low delay without prejudging the question of per-flow rate
policing.</t> policing.</t>
<t>Nonetheless, the interests of users or flows might conflict, <t>Nonetheless, the interests of users or flows might conflict,
e.g. in case of accident or malice. Then per-flow rate control e.g., in case of accident or malice. Then per-flow rate control
could be necessary. If flow-rate control is needed, it can be provided could be necessary. If per-flow rate control is needed, it can be provid
ed
as a modular addition to a DualQ. And similarly, if protection against as a modular addition to a DualQ. And similarly, if protection against
excessive queue delay is needed, a per-flow queue protection option excessive queue delay is needed, a per-flow queue protection option
can be added to a DualQ (e.g. <xref target="I-D.briscoe-docsis-q-protect ion" format="default"/>).</t> can be added to a DualQ (e.g., <xref target="I-D.briscoe-docsis-q-protec tion" format="default"/>).</t>
</section> </section>
<section anchor="dualq_Overload" numbered="true" toc="default"> <section anchor="dualq_Overload" numbered="true" toc="default">
<name>Handling Unresponsive Flows and Overload</name> <name>Handling Unresponsive Flows and Overload</name>
<t>In the absence of any per-flow control, it is important that the <t>In the absence of any per-flow control, it is important that the
basic DualQ Coupled AQM gives unresponsive flows no more throughput basic DualQ Coupled AQM gives unresponsive flows no more throughput
advantage than a single-queue AQM would, and that it at least handles advantage than a single-queue AQM would, and that it at least handles
overload situations. Overload means that incoming load significantly overload situations. Overload means that incoming load significantly
or persistently exceeds output capacity, but it is not intended to be or persistently exceeds output capacity, but it is not intended to be
a precise term -- significant and persistent are matters of a precise term -- significant and persistent are matters of
degree.</t> degree.</t>
<t>A trade-off needs to be made between complexity and the risk of <t>A trade-off needs to be made between complexity and the risk of
either traffic class harming the other. In overloaded conditions the either traffic class harming the other. In overloaded conditions, the
higher priority L4S service will have to sacrifice some aspect of its higher priority L4S service will have to sacrifice some aspect of its
performance. Depending on the degree of overload, alternative performance. Depending on the degree of overload, alternative
solutions may relax a different factor: e.g. throughput, delay, solutions may relax a different factor: for example, throughput, delay,
drop. These choices need to be made either by the developer or by or drop. These choices need to be made either by the developer or by
operator policy, rather than by the IETF. Subsequent subsections operator policy, rather than by the IETF.
discuss aspects relating to handling of different degrees of overload: Subsequent subsections
discuss handling different degrees of overload:
</t> </t>
<ul spacing="normal"> <ul spacing="normal">
<li> <li>
<t>Unresponsive flows (L and/or C) but not overloaded, <t>Unresponsive flows (L and/or C) but not overloaded,
i.e. the sum of unresponsive load before adding any i.e., the sum of unresponsive load before adding any
responsive traffic is below capacity;</t> responsive traffic is below capacity.</t>
<ul empty="true" spacing="normal"> <ul empty="true" spacing="normal">
<li>This case is handled by the regular Coupled DualQ (<xref targe t="dualq_coupled" format="default"/>) but not discussed there. So below, <li>This case is handled by the regular Coupled DualQ (<xref targe t="dualq_coupled" format="default"/>) but not discussed there. So below,
<xref target="dualq_unresponsive_wo_overload" format="default"/> explains the <xref target="dualq_unresponsive_wo_overload" format="default"/> explains the
design goal, and how it is achieved in practice;</li> design goal and how it is achieved in practice.</li>
</ul> </ul>
</li> </li>
<li> <li>
<t>Unresponsive flows (L and/or C) causing persistent overload, <t>Unresponsive flows (L and/or C) causing persistent overload,
i.e. the sum of unresponsive load even before adding any i.e., the sum of unresponsive load even before adding any
responsive traffic persistently exceeds capacity;</t> responsive traffic persistently exceeds capacity.</t>
<ul empty="true" spacing="normal"> <ul empty="true" spacing="normal">
<li>This case is not covered by the regular Coupled DualQ <li>This case is not covered by the regular Coupled DualQ
mechanism (<xref target="dualq_coupled" format="default"/>) but the last para mechanism (<xref target="dualq_coupled" format="default"/>), but the last paragraph
in <xref target="dualq_unexpected" format="default"/> sets out a requirement to in <xref target="dualq_unexpected" format="default"/> sets out a requirement to
handle the case where ECN-capable traffic could starve handle the case where ECN-capable traffic could starve
non-ECN-capable traffic. <xref target="dualq_Overload_Saturation " format="default"/> below discusses the non-ECN-capable traffic. <xref target="dualq_Overload_Saturation " format="default"/> below discusses the
general options and gives specific examples.</li> general options and gives specific examples.</li>
</ul> </ul>
</li> </li>
<li> <li>
<t>Short-term overload that lies between the 'not overloaded' and <t>Short-term overload that lies between the 'not overloaded' and
'persistently overloaded' cases. </t> 'persistently overloaded' cases.</t>
<ul empty="true" spacing="normal"> <ul empty="true" spacing="normal">
<li>For the period before overload is deemed persistent, <xref tar get="dualq_Overload_Starvation" format="default"/> discusses options for <li>For the period before overload is deemed persistent, <xref tar get="dualq_Overload_Starvation" format="default"/> discusses options for
more immediate mechanisms at the scheduler timescale. These more immediate mechanisms at the scheduler timescale. These
prevent short-term starvation of the C queue by making the prevent short-term starvation of the C queue by making the
priority of the L queue conditional, as required in <xref target ="dualq_functional_reqs" format="default"/>.</li> priority of the L queue conditional, as required in <xref target ="dualq_functional_reqs" format="default"/>.</li>
</ul> </ul>
</li> </li>
</ul> </ul>
<section anchor="dualq_unresponsive_wo_overload" numbered="true" toc="de fault"> <section anchor="dualq_unresponsive_wo_overload" numbered="true" toc="de fault">
<name>Unresponsive Traffic without Overload</name> <name>Unresponsive Traffic without Overload</name>
<t>When one or more L flows and/or C flows are unresponsive, but <t>When one or more L flows and/or C flows are unresponsive, but
their total load is within the link capacity so that they do not their total load is within the link capacity so that they do not
saturate the coupled marking (below 100%), the goal of a DualQ AQM saturate the coupled marking (below 100%), the goal of a DualQ AQM
is to behave no worse than a single-queue AQM.</t> is to behave no worse than a single-queue AQM.</t>
<t>Tests have shown that this is indeed the case with no additional <t>Tests have shown that this is indeed the case with no additional
mechanism beyond the regular Coupled DualQ of <xref target="dualq_coup led" format="default"/> (see the results of 'overload experiments' mechanism beyond the regular Coupled DualQ of <xref target="dualq_coup led" format="default"/> (see the results of 'overload experiments'
in <xref target="DCttH19" format="default"/>). Perhaps counter-intuiti vely, whether in <xref target="L4Seval22" format="default"/>). Perhaps counterintuit ively, whether
the unresponsive flow classifies itself into the L or the C queue, the unresponsive flow classifies itself into the L or the C queue,
the DualQ system behaves as if it has subtracted from the overall the DualQ system behaves as if it has subtracted from the overall
link capacity. Then, the coupling shares out the remaining capacity link capacity. Then, the coupling shares out the remaining capacity
between any competing responsive flows (in either queue). See also between any competing responsive flows (in either queue). See also
<xref target="dualq_Overload_Starvation" format="default"/>, which dis cusses <xref target="dualq_Overload_Starvation" format="default"/>, which dis cusses
scheduler-specific details.</t> scheduler-specific details.</t>
</section> </section>
<section anchor="dualq_Overload_Starvation" numbered="true" toc="default "> <section anchor="dualq_Overload_Starvation" numbered="true" toc="default ">
<name>Avoiding Short-Term Classic Starvation: Sacrifice L4S Throughput or Delay?</name> <name>Avoiding Short-Term Classic Starvation: Sacrifice L4S Throughput or Delay?</name>
<t>Priority of L4S is required to be conditional (see <xref target="du alq_coupled_structure" format="default"/> &amp; <xref target="dualq_functional_r eqs" format="default"/>) to avoid short-term starvation of <t>Priority of L4S is required to be conditional (see Sections <xref t arget="dualq_coupled_structure" format="counter"/> and <xref target="dualq_funct ional_reqs" format="counter"/>) to avoid short-term starvation of
Classic. Otherwise, as explained in <xref target="dualq_coupled_struct ure" format="default"/>, even a lone responsive L4S flow Classic. Otherwise, as explained in <xref target="dualq_coupled_struct ure" format="default"/>, even a lone responsive L4S flow
could temporarily block a small finite set of C packets could temporarily block a small finite set of C packets
(e.g. an initial window or DNS request). The blockage would (e.g., an initial window or DNS request). The blockage would
only be brief, but it could be longer for certain AQM only be brief, but it could be longer for certain AQM
implementations that can only increase the congestion signal coupled implementations that can only increase the congestion signal coupled
from the C queue when C packets are actually being dequeued. There from the C queue when C packets are actually being dequeued. There
is then the question of whether to sacrifice L4S throughput or L4S is then the question of whether to sacrifice L4S throughput or L4S
delay (or some other policy) to make the priority conditional:</t> delay (or some other policy) to make the priority conditional:</t>
<dl newline="false" spacing="normal"> <dl newline="true" spacing="normal">
<dt>Sacrifice L4S throughput: </dt> <dt>Sacrifice L4S throughput: </dt>
<dd anchor="dualq_Minimum_Service"> <dd anchor="dualq_Minimum_Service">
<t>By using weighted <t>By using WRR as the conditional priority scheduler, the L4S
round-robin as the conditional priority scheduler, the L4S
service can sacrifice some throughput during overload. This can service can sacrifice some throughput during overload. This can
either be thought of as guaranteeing a minimum throughput be thought of as guaranteeing either a minimum throughput
service for Classic traffic, or as guaranteeing a maximum delay service for Classic traffic or a maximum delay
for a packet at the head of the Classic queue.</t> for a packet at the head of the Classic queue.</t>
<t>Cautionary note: a WRR scheduler can only <aside><t>Cautionary note: a WRR scheduler can only
guarantee Classic throughput if Classic sources are sending guarantee Classic throughput if Classic sources are sending
enough to use it -- congestion signals can undermine enough to use it -- congestion signals can undermine
scheduling because they determine how much responsive traffic of scheduling because they determine how much responsive traffic of
each class arrives for scheduling in the first place. This is each class arrives for scheduling in the first place. This is
why scheduling is only relied on to handle short-term why scheduling is only relied on to handle short-term
starvation; until congestion signals build up and the sources starvation, until congestion signals build up and the sources
react. Even during long-term overload (discussed more fully in react. Even during long-term overload (discussed more fully in
<xref target="dualq_Overload_Saturation" format="default"/>), it's pragmatic to <xref target="dualq_Overload_Saturation" format="default"/>), it's pragmatic to
discard packets from both queues, which again thins the traffic discard packets from both queues, which again thins the traffic
before it reaches the scheduler. This is because a scheduler before it reaches the scheduler. This is because a scheduler
cannot be relied on to handle long-term overload since the right cannot be relied on to handle long-term overload since the right
scheduler weight cannot be known for every scenario.</t> scheduler weight cannot be known for every scenario.</t></aside>
<t>The scheduling weight of the Classic queue <t>The scheduling weight of the Classic queue
should be small (e.g. 1/16). In most traffic scenarios the should be small (e.g., 1/16). In most traffic scenarios, the
scheduler will not interfere and it will not need to, because scheduler will not interfere and it will not need to, because
the coupling mechanism and the end-systems will determine the the coupling mechanism and the end systems will determine the
share of capacity across both queues as if it were a single share of capacity across both queues as if it were a single
pool. However, if L4S traffic is over-aggressive or pool. However, if L4S traffic is over-aggressive or
unresponsive, the scheduler weight for Classic traffic will at unresponsive, the scheduler weight for Classic traffic will at
least be large enough to ensure it does not starve in the least be large enough to ensure it does not starve in the
short-term. </t> short term. </t>
<t>Although WRR scheduling is <t>Although WRR scheduling is
only expected to address short-term overload, there are only expected to address short-term overload, there are
(somewhat rare) cases when WRR has an effect on capacity shares (somewhat rare) cases when WRR has an effect on capacity shares
over longer time-scales. But its effect is minor, and it over longer timescales. But its effect is minor, and it
certainly does no harm. Specifically, in cases where the ratio certainly does no harm. Specifically, in cases where the ratio
of L4S to Classic flows (e.g. 19:1) is greater than the of L4S to Classic flows (e.g., 19:1) is greater than the
ratio of their scheduler weights (e.g. 15:1), the L4S flows ratio of their scheduler weights (e.g., 15:1), the L4S flows
will get less than an equal share of the capacity, but only will get less than an equal share of the capacity, but only
slightly. For instance, with the example numbers given, each L4S slightly. For instance, with the example numbers given, each L4S
flow will get (15/16)/19 = 4.9% when ideally each would get flow will get (15/16)/19 = 4.9% when ideally each would get
1/20=5%. In the rather specific case of an unresponsive flow 1/20 = 5%. In the rather specific case of an unresponsive flow
taking up just less than the capacity set aside for L4S taking up just less than the capacity set aside for L4S
(e.g. 14/16 in the above example), using WRR could (e.g., 14/16 in the above example), using WRR could
significantly reduce the capacity left for any responsive L4S significantly reduce the capacity left for any responsive L4S
flows.</t> flows.</t>
<t>The scheduling weight of the <t>The scheduling weight of the
Classic queue should not be too small, otherwise a C packet at Classic queue should not be too small, otherwise a C packet at
the head of the queue could be excessively delayed by a the head of the queue could be excessively delayed by a
continually busy L queue. For instance if the Classic weight is continually busy L queue. For instance, if the Classic weight is
1/16, the maximum that a Classic packet at the head of the queue 1/16, the maximum that a Classic packet at the head of the queue
can be delayed by L traffic is the serialization delay of 15 can be delayed by L traffic is the serialization delay of 15
MTU-sized packets.</t> MTU-sized packets.</t>
</dd> </dd>
<dt>Sacrifice L4S Delay:</dt> <dt>Sacrifice L4S delay:</dt>
<dd anchor="dualq_Delay_Overload"> <dd anchor="dualq_Delay_Overload">
<t>The operator could choose to <t>The operator could choose to
control overload of the Classic queue by allowing some delay to control overload of the Classic queue by allowing some delay to
'leak' across to the L4S queue. The scheduler can be made to 'leak' across to the L4S queue. The scheduler can be made to
behave like a single First-In First-Out (FIFO) queue with behave like a single FIFO queue with
different service times by implementing a very simple different service times by implementing a very simple
conditional priority scheduler that could be called a conditional priority scheduler that could be called a
"time-shifted FIFO" (see the Modifier Earliest Deadline First "time-shifted FIFO" (TS-FIFO) (see the Modifier Earliest Deadline
(MEDF) scheduler <xref target="MEDF" format="default"/>). This sch First
eduler (MEDF) scheduler <xref target="MEDF" format="default"/>). This sch
eduler
adds tshift to the queue delay of the next L4S packet, before adds tshift to the queue delay of the next L4S packet, before
comparing it with the queue delay of the next Classic packet, comparing it with the queue delay of the next Classic packet,
then it selects the packet with the greater adjusted queue then it selects the packet with the greater adjusted queue
delay.</t> delay.</t>
<t>Under regular conditions, this <t>Under regular conditions, the
time-shifted FIFO scheduler behaves just like a strict priority TS-FIFO scheduler behaves just like a strict priority
scheduler. But under moderate or high overload it prevents scheduler. But under moderate or high overload, it prevents
starvation of the Classic queue, because the time-shift (tshift) starvation of the Classic queue, because the time-shift (tshift)
defines the maximum extra queuing delay of Classic packets defines the maximum extra queuing delay of Classic packets
relative to L4S. This would control milder overload of relative to L4S.
This would control milder overload of
responsive traffic by introducing delay to defer invoking the responsive traffic by introducing delay to defer invoking the
overload mechanisms in <xref target="dualq_Overload_Saturation" fo rmat="default"/>, particularly when close to overload mechanisms in <xref target="dualq_Overload_Saturation" fo rmat="default"/>, particularly when close to
the maximum congestion signal.</t> the maximum congestion signal.</t>
</dd> </dd>
</dl> </dl>
<t>The example implementations in <xref target="dualq_Ex_algo_pi2" for <t>The example implementations in Appendices <xref target="dualq_Ex_al
mat="default"/> go_pi2" format="counter"/>
and <xref target="dualq_Ex_algo" format="default"/> could both be impl and <xref target="dualq_Ex_algo" format="counter"/> could both be impl
emented with emented with
either policy.</t> either policy.</t>
</section> </section>
<section anchor="dualq_Overload_Saturation" numbered="true" toc="default "> <section anchor="dualq_Overload_Saturation" numbered="true" toc="default ">
<name>L4S ECN Saturation: Introduce Drop or Delay?</name> <name>L4S ECN Saturation: Introduce Drop or Delay?</name>
<t>This section concerns persistent overload caused by unresponsive <t>This section concerns persistent overload caused by unresponsive
L and/or C flows. To keep the throughput of both L4S and Classic L and/or C flows. To keep the throughput of both L4S and Classic
flows roughly equal over the full load range, a different control flows roughly equal over the full load range, a different control
strategy needs to be defined above the point where the L4S AQM strategy needs to be defined above the point where the L4S AQM
persistently saturates to an ECN marking probability of 100% leaving persistently saturates to an ECN marking probability of 100%, leaving
no room to push back the load any harder. L4S ECN marking will no room to push back the load any harder. L4S ECN marking will
saturate first (assuming the coupling factor k&gt;1), even though saturate first (assuming the coupling factor k&gt;1), even though
saturation could be caused by the sum of unresponsive traffic in saturation could be caused by the sum of unresponsive traffic in
either or both queues exceeding the link capacity.</t> either or both queues exceeding the link capacity.</t>
<t>The term 'unresponsive' includes cases where a flow becomes <t>The term 'unresponsive' includes cases where a flow becomes
temporarily unresponsive, for instance, a real-time flow that takes temporarily unresponsive, for instance, a real-time flow that takes
a while to adapt its rate in response to congestion, or a standard a while to adapt its rate in response to congestion, or a standard
Reno flow that is normally responsive, but above a certain Reno flow that is normally responsive, but above a certain
congestion level it will not be able to reduce its congestion window congestion level it will not be able to reduce its congestion window
below the allowed minimum of 2 segments <xref target="RFC5681" format= below the allowed minimum of 2 segments <xref target="RFC5681" format=
"default"/>, effectively becoming unresponsive. (Note that "default"/>, effectively becoming unresponsive. (Note that
L4S traffic ought to remain responsive below a window of 2 segments L4S traffic ought to remain responsive below a window of 2 segments.
(see the L4S requirements <xref target="I-D.ietf-tsvwg-ecn-l4s-id" for See the L4S requirements <xref target="RFC9331" format="default"/>.)</
mat="default"/>).</t> t>
<t>Saturation raises the question of whether to relieve congestion <t>Saturation raises the question of whether to relieve congestion
by introducing some drop into the L4S queue or by allowing delay to by introducing some drop into the L4S queue or by allowing delay to
grow in both queues (which could eventually lead to drop due to grow in both queues (which could eventually lead to drop due to
buffer exhaustion anyway):</t> buffer exhaustion anyway):</t>
<dl newline="false" spacing="normal"> <dl newline="true" spacing="normal">
<dt>Drop on Saturation:</dt> <dt>Drop on Saturation:</dt>
<dd>Persistent saturation can be <dd>Persistent saturation can be
defined by a maximum threshold for coupled L4S ECN marking defined by a maximum threshold for coupled L4S ECN marking
(assuming k&gt;1) before saturation starts to make the flow (assuming k&gt;1) before saturation starts to make the flow
rates of the different traffic types diverge. Above that, the rates of the different traffic types diverge. Above that, the
drop probability of Classic traffic is applied to all packets of drop probability of Classic traffic is applied to all packets of
all traffic types. Then experiments have shown that queueing all traffic types. Then experiments have shown that queuing
delay can be kept at the target in any overload situation, delay can be kept at the target in any overload situation,
including with unresponsive traffic, and no further measures are including with unresponsive traffic, and no further measures are
required (<xref target="dualq_overload_unresp_ect" format="default "/>).</dd> required (<xref target="dualq_overload_unresp_ect" format="default "/>).</dd>
<dt>Delay on Saturation:</dt> <dt>Delay on Saturation:</dt>
<dd>When L4S marking saturates, <dd>When L4S marking saturates,
instead of introducing L4S drop, the drop and marking instead of introducing L4S drop, the drop and marking
probabilities of both queues could be capped. Beyond that, delay probabilities of both queues could be capped. Beyond that, delay
will grow either solely in the queue with unresponsive traffic will grow either solely in the queue with unresponsive traffic
(if WRR is used), or in both queues (if time-shifted FIFO is (if WRR is used) or in both queues (if TS-FIFO is
used). In either case, the higher delay ought to control used). In either case, the higher delay ought to control
temporary high congestion. If the overload is more persistent, temporary high congestion. If the overload is more persistent,
eventually the combined DualQ will overflow and tail drop will eventually the combined DualQ will overflow and tail drop will
control congestion.</dd> control congestion.</dd>
</dl> </dl>
<t>The example implementation in <xref target="dualq_Ex_algo_pi2" form at="default"/> <t>The example implementation in <xref target="dualq_Ex_algo_pi2" form at="default"/>
solely applies the "drop on saturation" policy. The DOCSIS solely applies the "drop on saturation" policy. The DOCSIS
specification of a DualQ Coupled AQM <xref target="DOCSIS3.1" format=" default"/> specification of a DualQ Coupled AQM <xref target="DOCSIS3.1" format=" default"/>
also implements the 'drop on saturation' policy with a very shallow also implements the 'drop on saturation' policy with a very shallow
L buffer. However, the addition of DOCSIS per-flow Queue Protection L buffer. However, the addition of DOCSIS per-flow Queue Protection
<xref target="I-D.briscoe-docsis-q-protection" format="default"/> turn s this into <xref target="I-D.briscoe-docsis-q-protection" format="default"/> turn s this into
'delay on saturation' by redirecting some packets of the flow(s) 'delay on saturation' by redirecting some packets of the flow or flows
most responsible for L queue overload into the C queue, which has a that are most responsible for L queue overload into the C queue, which
has a
higher delay target. If overload continues, this again becomes 'drop higher delay target. If overload continues, this again becomes 'drop
on saturation' as the level of drop in the C queue rises to maintain on saturation' as the level of drop in the C queue rises to maintain
the target delay of the C queue.</t> the target delay of the C queue.</t>
<section anchor="dualq_overload_unresp_ect" numbered="true" toc="defau lt"> <section anchor="dualq_overload_unresp_ect" numbered="true" toc="defau lt">
<name>Protecting against Overload by Unresponsive ECN-Capable Traffi c</name> <name>Protecting against Overload by Unresponsive ECN-Capable Traffi c</name>
<t>Without a specific overload mechanism, unresponsive traffic <t>Without a specific overload mechanism, unresponsive traffic
would have a greater advantage if it were also ECN-capable. The would have a greater advantage if it were also ECN-capable. The
advantage is undetectable at normal low levels of marking. advantage is undetectable at normal low levels of marking.
However, it would become significant with the higher levels of However, it would become significant with the higher levels of
marking typical during overload, when it could evade a significant marking typical during overload, when it could evade a significant
degree of drop. This is an issue whether the ECN-capable traffic degree of drop. This is an issue whether the ECN-capable traffic
is L4S or Classic.</t> is L4S or Classic.</t>
<t>This raises the question of whether and when to introduce drop <t>This raises the question of whether and when to introduce drop
of ECN-capable traffic, as required by both Section 7 of the ECN of ECN-capable traffic, as required by both Section <xref target="RF
spec <xref target="RFC3168" format="default"/> and Section 4.2.1 of C3168" sectionFormat="bare" section="7"/> of the ECN spec <xref target="RFC3168"
the AQM format="default"/> and Section <xref target="RFC7567" sectionFormat="bare" sect
recommendations <xref target="RFC7567" format="default"/>.</t> ion="4.2.1"/> of the AQM
recommendations <xref target="RFC7567" format="default"/>.</t>
<t>As an example, experiments with the DualPI2 AQM (<xref target="du alq_Ex_algo_pi2" format="default"/>) have shown that introducing 'drop on <t>As an example, experiments with the DualPI2 AQM (<xref target="du alq_Ex_algo_pi2" format="default"/>) have shown that introducing 'drop on
saturation' at 100% coupled L4S marking addresses this problem saturation' at 100% coupled L4S marking addresses this problem
with unresponsive ECN as well as addressing the saturation with unresponsive ECN, and it also addresses the saturation
problem. At saturation, DualPI2 switches into overload mode, where problem. At saturation, DualPI2 switches into overload mode, where
the base AQM is driven by the max delay of both queues and it the Base AQM is driven by the max delay of both queues, and it
introduces probabilistic drop to both queues equally. It leaves introduces probabilistic drop to both queues equally.
It leaves
only a small range of congestion levels just below saturation only a small range of congestion levels just below saturation
where unresponsive traffic gains any advantage from using the ECN where unresponsive traffic gains any advantage from using the ECN
capability (relative to being unresponsive without ECN), and the capability (relative to being unresponsive without ECN), and the
advantage is hardly detectable (see <xref target="DualQ-Test" format ="default"/> advantage is hardly detectable (see <xref target="DualQ-Test" format ="default"/>
and section IV-E of <xref target="DCttH19" format="default"/>. Also overload with and section IV-G of <xref target="L4Seval22" format="default"/>). Al so, overload with
an unresponsive ECT(1) flow gets no more bandwidth advantage than an unresponsive ECT(1) flow gets no more bandwidth advantage than
with ECT(0).</t> with ECT(0).</t>
</section> </section>
</section> </section>
</section> </section>
</section> </section>
</middle> </middle>
<!-- *****BACK MATTER ***** -->
<back> <back>
<displayreference target="I-D.briscoe-tsvwg-l4s-diffserv" to="L4S-DIFFSERV"/>
<displayreference target="I-D.briscoe-docsis-q-protection" to="DOCSIS-Q-PROT"/>
<displayreference target="I-D.cardwell-iccrg-bbr-congestion-control" to="BBR-CC"
/>
<displayreference target="I-D.briscoe-iccrg-prague-congestion-control" to="PRAGU
E-CC"/>
<displayreference target="I-D.mathis-iccrg-relentless-tcp" to="RELENTLESS"/>
<references> <references>
<name>References</name> <name>References</name>
<references> <references>
<name>Normative References</name> <name>Normative References</name>
<reference anchor="RFC2119" target="https://www.rfc-editor.org/info/rfc2
119" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.2119.xml"> <xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.2119.
<front> xml"/>
<title>Key words for use in RFCs to Indicate Requirement Levels</tit <xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.3168.
le> xml"/>
<author initials="S." surname="Bradner" fullname="S. Bradner"> <xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8311.
<organization/> xml"/>
</author>
<date year="1997" month="March"/> <!-- [I-D.ietf-tsvwg-ecn-l4s-id] companion doc 9331 - title matches as of 1/17/2
<abstract> 3-->
<t>In many standards track documents several words are used to sig <reference anchor='RFC9331' target='https://www.rfc-editor.org/info/rfc9331'>
nify the requirements in the specification. These words are often capitalized. <front>
This document defines these words as they should be interpreted in IETF document <title>The Explicit Congestion Notification (ECN) Protocol for Low Latency, Low
s. This document specifies an Internet Best Current Practices for the Internet Loss, and Scalable Throughput (L4S)</title>
Community, and requests discussion and suggestions for improvements.</t> <author initials='K' surname='De Schepper' fullname='Koen De Schepper'>
</abstract> <organization />
</front> </author>
<seriesInfo name="BCP" value="14"/> <author initials='B' surname='Briscoe' fullname='Bob Briscoe' role='editor'>
<seriesInfo name="RFC" value="2119"/> <organization />
<seriesInfo name="DOI" value="10.17487/RFC2119"/> </author>
</reference> <date month='January' year='2023'/>
<reference anchor="RFC3168" target="https://www.rfc-editor.org/info/rfc3 </front>
168" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.3168.xml"> <seriesInfo name="RFC" value="9331"/>
<front> <seriesInfo name="DOI" value="10.17487/RFC9331"/>
<title>The Addition of Explicit Congestion Notification (ECN) to IP< </reference>
/title>
<author initials="K." surname="Ramakrishnan" fullname="K. Ramakrishn
an">
<organization/>
</author>
<author initials="S." surname="Floyd" fullname="S. Floyd">
<organization/>
</author>
<author initials="D." surname="Black" fullname="D. Black">
<organization/>
</author>
<date year="2001" month="September"/>
<abstract>
<t>This memo specifies the incorporation of ECN (Explicit Congesti
on Notification) to TCP and IP, including ECN's use of two bits in the IP header
. [STANDARDS-TRACK]</t>
</abstract>
</front>
<seriesInfo name="RFC" value="3168"/>
<seriesInfo name="DOI" value="10.17487/RFC3168"/>
</reference>
<reference anchor="RFC8311" target="https://www.rfc-editor.org/info/rfc8
311" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8311.xml">
<front>
<title>Relaxing Restrictions on Explicit Congestion Notification (EC
N) Experimentation</title>
<author initials="D." surname="Black" fullname="D. Black">
<organization/>
</author>
<date year="2018" month="January"/>
<abstract>
<t>This memo updates RFC 3168, which specifies Explicit Congestion
Notification (ECN) as an alternative to packet drops for indicating network con
gestion to endpoints. It relaxes restrictions in RFC 3168 that hinder experimen
tation towards benefits beyond just removal of loss. This memo summarizes the a
nticipated areas of experimentation and updates RFC 3168 to enable experimentati
on in these areas. An Experimental RFC in the IETF document stream is required
to take advantage of any of these enabling updates. In addition, this memo make
s related updates to the ECN specifications for RTP in RFC 6679 and for the Data
gram Congestion Control Protocol (DCCP) in RFCs 4341, 4342, and 5622. This memo
also records the conclusion of the ECN nonce experiment in RFC 3540 and provide
s the rationale for reclassification of RFC 3540 from Experimental to Historic;
this reclassification enables new experimental use of the ECT(1) codepoint.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="8311"/>
<seriesInfo name="DOI" value="10.17487/RFC8311"/>
</reference>
<reference anchor="I-D.ietf-tsvwg-ecn-l4s-id" target="https://datatracke
r.ietf.org/api/v1/doc/document/draft-ietf-tsvwg-ecn-l4s-id/" xml:base="https://b
ib.ietf.org/public/rfc/bibxml-ids/reference.I-D.ietf-tsvwg-ecn-l4s-id.xml">
<front>
<title>Explicit Congestion Notification (ECN) Protocol for Very Low
Queuing Delay (L4S)</title>
<author fullname="Koen De Schepper"/>
<author fullname="Bob Briscoe"/>
<date day="8" month="August" year="2022"/>
<abstract>
<t>This specification defines the protocol to be used for a new ne
twork
service called low latency, low loss and scalable throughput (L4S).
L4S uses an Explicit Congestion Notification (ECN) scheme at the IP
layer that is similar to the original (or 'Classic') ECN approach,
except as specified within. L4S uses 'scalable' congestion control,
which induces much more frequent control signals from the network and
it responds to them with much more fine-grained adjustments, so that
very low (typically sub-millisecond on average) and consistently low
queuing delay becomes possible for L4S traffic without compromising
link utilization. Thus even capacity-seeking (TCP-like) traffic can
have high bandwidth and very low delay at the same time, even during
periods of high traffic load.</t>
<t>The L4S identifier defined in this document distinguishes L4S f
rom
'Classic' (e.g. TCP-Reno-friendly) traffic. Then, network
bottlenecks can be incrementally modified to distinguish and isolate
existing traffic that still follows the Classic behaviour, to prevent
it degrading the low queuing delay and low loss of L4S traffic. This
experimental track specification defines the rules that L4S
transports and network elements need to follow, with the intention
that L4S flows neither harm each other's performance nor that of
Classic traffic. It also suggests open questions to be investigated
during experimentation. Examples of new active queue management
(AQM) marking algorithms and examples of new transports (whether TCP-
like or real-time) are specified separately.</t>
</abstract>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-tsvwg-ecn-l4s-id-2
8"/>
</reference>
</references> </references>
<references> <references>
<name>Informative References</name> <name>Informative References</name>
<reference anchor="RFC0970" target="https://www.rfc-editor.org/info/rfc9
70" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.0970.xml"> <xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.0970.
<front> xml"/>
<title>On Packet Switches With Infinite Storage</title>
<author initials="J." surname="Nagle" fullname="J. Nagle"> <reference anchor="RED" target="https://dl.acm.org/doi/10.1109/90.251892
<organization/> ">
</author>
<date year="1985" month="December"/>
<abstract>
<t>The purpose of this RFC is to focus discussion on a particular
problem in the ARPA-Internet and possible methods of solution. Most prior wo
rk on congestion in datagram systems focuses on buffer management. In this
memo the case of a packet switch with infinite storage is considered. Such
a packet switch can never run out of buffers. It can, however, still become
congested. The meaning of congestion in an infinite-storage system is explor
ed. An unexpected result is found that shows a datagram network with infinit
e storage, first-in-first-out queuing, at least two packet switches, and a fi
nite packet lifetime will, under overload, drop all packets. By attacking th
e problem of congestion for the infinite-storage case, new solutions applicab
le to switches with finite storage may be found. No proposed solutions this
document are intended as standards for the ARPA-Internet at this time.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="970"/>
<seriesInfo name="DOI" value="10.17487/RFC0970"/>
</reference>
<reference anchor="RFC2309" target="https://www.rfc-editor.org/info/rfc2
309" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.2309.xml">
<front>
<title>Recommendations on Queue Management and Congestion Avoidance
in the Internet</title>
<author initials="B." surname="Braden" fullname="B. Braden">
<organization/>
</author>
<author initials="D." surname="Clark" fullname="D. Clark">
<organization/>
</author>
<author initials="J." surname="Crowcroft" fullname="J. Crowcroft">
<organization/>
</author>
<author initials="B." surname="Davie" fullname="B. Davie">
<organization/>
</author>
<author initials="S." surname="Deering" fullname="S. Deering">
<organization/>
</author>
<author initials="D." surname="Estrin" fullname="D. Estrin">
<organization/>
</author>
<author initials="S." surname="Floyd" fullname="S. Floyd">
<organization/>
</author>
<author initials="V." surname="Jacobson" fullname="V. Jacobson">
<organization/>
</author>
<author initials="G." surname="Minshall" fullname="G. Minshall">
<organization/>
</author>
<author initials="C." surname="Partridge" fullname="C. Partridge">
<organization/>
</author>
<author initials="L." surname="Peterson" fullname="L. Peterson">
<organization/>
</author>
<author initials="K." surname="Ramakrishnan" fullname="K. Ramakrishn
an">
<organization/>
</author>
<author initials="S." surname="Shenker" fullname="S. Shenker">
<organization/>
</author>
<author initials="J." surname="Wroclawski" fullname="J. Wroclawski">
<organization/>
</author>
<author initials="L." surname="Zhang" fullname="L. Zhang">
<organization/>
</author>
<date year="1998" month="April"/>
<abstract>
<t>This memo presents two recommendations to the Internet communit
y concerning measures to improve and preserve Internet performance. It presents
a strong recommendation for testing, standardization, and widespread deployment
of active queue management in routers, to improve the performance of today's In
ternet. It also urges a concerted effort of research, measurement, and ultimate
deployment of router mechanisms to protect the Internet from flows that are not
sufficiently responsive to congestion notification. This memo provides informa
tion for the Internet community. It does not specify an Internet standard of an
y kind.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="2309"/>
<seriesInfo name="DOI" value="10.17487/RFC2309"/>
</reference>
<reference anchor="RFC2914" target="https://www.rfc-editor.org/info/rfc2
914" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.2914.xml">
<front>
<title>Congestion Control Principles</title>
<author initials="S." surname="Floyd" fullname="S. Floyd">
<organization/>
</author>
<date year="2000" month="September"/>
<abstract>
<t>The goal of this document is to explain the need for congestion
control in the Internet, and to discuss what constitutes correct congestion con
trol. This document specifies an Internet Best Current Practices for the Intern
et Community, and requests discussion and suggestions for improvements.</t>
</abstract>
</front>
<seriesInfo name="BCP" value="41"/>
<seriesInfo name="RFC" value="2914"/>
<seriesInfo name="DOI" value="10.17487/RFC2914"/>
</reference>
<reference anchor="RFC3246" target="https://www.rfc-editor.org/info/rfc3
246" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.3246.xml">
<front>
<title>An Expedited Forwarding PHB (Per-Hop Behavior)</title>
<author initials="B." surname="Davie" fullname="B. Davie">
<organization/>
</author>
<author initials="A." surname="Charny" fullname="A. Charny">
<organization/>
</author>
<author initials="J.C.R." surname="Bennet" fullname="J.C.R. Bennet">
<organization/>
</author>
<author initials="K." surname="Benson" fullname="K. Benson">
<organization/>
</author>
<author initials="J.Y." surname="Le Boudec" fullname="J.Y. Le Boudec
">
<organization/>
</author>
<author initials="W." surname="Courtney" fullname="W. Courtney">
<organization/>
</author>
<author initials="S." surname="Davari" fullname="S. Davari">
<organization/>
</author>
<author initials="V." surname="Firoiu" fullname="V. Firoiu">
<organization/>
</author>
<author initials="D." surname="Stiliadis" fullname="D. Stiliadis">
<organization/>
</author>
<date year="2002" month="March"/>
<abstract>
<t>This document defines a PHB (per-hop behavior) called Expedited
Forwarding (EF). The PHB is a basic building block in the Differentiated Servi
ces architecture. EF is intended to provide a building block for low delay, low
jitter and low loss services by ensuring that the EF aggregate is served at a c
ertain configured rate. This document obsoletes RFC 2598. [STANDARDS-TRACK]</t>
</abstract>
</front>
<seriesInfo name="RFC" value="3246"/>
<seriesInfo name="DOI" value="10.17487/RFC3246"/>
</reference>
<reference anchor="RFC3649" target="https://www.rfc-editor.org/info/rfc3
649" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.3649.xml">
<front>
<title>HighSpeed TCP for Large Congestion Windows</title>
<author initials="S." surname="Floyd" fullname="S. Floyd">
<organization/>
</author>
<date year="2003" month="December"/>
<abstract>
<t>The proposals in this document are experimental. While they ma
y be deployed in the current Internet, they do not represent a consensus that th
is is the best method for high-speed congestion control. In particular, we note
that alternative experimental proposals are likely to be forthcoming, and it is
not well understood how the proposals in this document will interact with such
alternative proposals. This document proposes HighSpeed TCP, a modification to
TCP's congestion control mechanism for use with TCP connections with large conge
stion windows. The congestion control mechanisms of the current Standard TCP co
nstrains the congestion windows that can be achieved by TCP in realistic environ
ments. For example, for a Standard TCP connection with 1500-byte packets and a
100 ms round-trip time, achieving a steady-state throughput of 10 Gbps would req
uire an average congestion window of 83,333 segments, and a packet drop rate of
at most one congestion event every 5,000,000,000 packets (or equivalently, at mo
st one congestion event every 1 2/3 hours). This is widely acknowledged as an u
nrealistic constraint. To address his limitation of TCP, this document proposes
HighSpeed TCP, and solicits experimentation and feedback from the wider communi
ty.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="3649"/>
<seriesInfo name="DOI" value="10.17487/RFC3649"/>
</reference>
<reference anchor="RFC5033" target="https://www.rfc-editor.org/info/rfc5
033" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.5033.xml">
<front>
<title>Specifying New Congestion Control Algorithms</title>
<author initials="S." surname="Floyd" fullname="S. Floyd">
<organization/>
</author>
<author initials="M." surname="Allman" fullname="M. Allman">
<organization/>
</author>
<date year="2007" month="August"/>
<abstract>
<t>The IETF's standard congestion control schemes have been widely
shown to be inadequate for various environments (e.g., high-speed networks). R
ecent research has yielded many alternate congestion control schemes that signif
icantly differ from the IETF's congestion control principles. Using these new c
ongestion control schemes in the global Internet has possible ramifications to b
oth the traffic using the new congestion control and to traffic using the curren
tly standardized congestion control. Therefore, the IETF must proceed with caut
ion when dealing with alternate congestion control proposals. The goal of this
document is to provide guidance for considering alternate congestion control alg
orithms within the IETF. This document specifies an Internet Best Current Pract
ices for the Internet Community, and requests discussion and suggestions for imp
rovements.</t>
</abstract>
</front>
<seriesInfo name="BCP" value="133"/>
<seriesInfo name="RFC" value="5033"/>
<seriesInfo name="DOI" value="10.17487/RFC5033"/>
</reference>
<reference anchor="RFC5348" target="https://www.rfc-editor.org/info/rfc5
348" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.5348.xml">
<front>
<title>TCP Friendly Rate Control (TFRC): Protocol Specification</tit
le>
<author initials="S." surname="Floyd" fullname="S. Floyd">
<organization/>
</author>
<author initials="M." surname="Handley" fullname="M. Handley">
<organization/>
</author>
<author initials="J." surname="Padhye" fullname="J. Padhye">
<organization/>
</author>
<author initials="J." surname="Widmer" fullname="J. Widmer">
<organization/>
</author>
<date year="2008" month="September"/>
<abstract>
<t>This document specifies TCP Friendly Rate Control (TFRC). TFRC
is a congestion control mechanism for unicast flows operating in a best-effort
Internet environment. It is reasonably fair when competing for bandwidth with T
CP flows, but has a much lower variation of throughput over time compared with T
CP, making it more suitable for applications such as streaming media where a rel
atively smooth sending rate is of importance.</t>
<t>This document obsoletes RFC 3448 and updates RFC 4342. [STANDA
RDS-TRACK]</t>
</abstract>
</front>
<seriesInfo name="RFC" value="5348"/>
<seriesInfo name="DOI" value="10.17487/RFC5348"/>
</reference>
<reference anchor="RFC5681" target="https://www.rfc-editor.org/info/rfc5
681" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.5681.xml">
<front>
<title>TCP Congestion Control</title>
<author initials="M." surname="Allman" fullname="M. Allman">
<organization/>
</author>
<author initials="V." surname="Paxson" fullname="V. Paxson">
<organization/>
</author>
<author initials="E." surname="Blanton" fullname="E. Blanton">
<organization/>
</author>
<date year="2009" month="September"/>
<abstract>
<t>This document defines TCP's four intertwined congestion control
algorithms: slow start, congestion avoidance, fast retransmit, and fast recover
y. In addition, the document specifies how TCP should begin transmission after
a relatively long idle period, as well as discussing various acknowledgment gene
ration methods. This document obsoletes RFC 2581. [STANDARDS-TRACK]</t>
</abstract>
</front>
<seriesInfo name="RFC" value="5681"/>
<seriesInfo name="DOI" value="10.17487/RFC5681"/>
</reference>
<reference anchor="RFC5706" target="https://www.rfc-editor.org/info/rfc5
706" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.5706.xml">
<front>
<title>Guidelines for Considering Operations and Management of New P
rotocols and Protocol Extensions</title>
<author initials="D." surname="Harrington" fullname="D. Harrington">
<organization/>
</author>
<date year="2009" month="November"/>
<abstract>
<t>New protocols or protocol extensions are best designed with due
consideration of the functionality needed to operate and manage the protocols.
Retrofitting operations and management is sub-optimal. The purpose of this docu
ment is to provide guidance to authors and reviewers of documents that define ne
w protocols or protocol extensions regarding aspects of operations and managemen
t that should be considered. This memo provides information for the Internet co
mmunity.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="5706"/>
<seriesInfo name="DOI" value="10.17487/RFC5706"/>
</reference>
<reference anchor="RFC7567" target="https://www.rfc-editor.org/info/rfc7
567" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.7567.xml">
<front>
<title>IETF Recommendations Regarding Active Queue Management</title
>
<author initials="F." surname="Baker" fullname="F. Baker" role="edit
or">
<organization/>
</author>
<author initials="G." surname="Fairhurst" fullname="G. Fairhurst" ro
le="editor">
<organization/>
</author>
<date year="2015" month="July"/>
<abstract>
<t>This memo presents recommendations to the Internet community co
ncerning measures to improve and preserve Internet performance. It presents a s
trong recommendation for testing, standardization, and widespread deployment of
active queue management (AQM) in network devices to improve the performance of t
oday's Internet. It also urges a concerted effort of research, measurement, and
ultimate deployment of AQM mechanisms to protect the Internet from flows that a
re not sufficiently responsive to congestion notification.</t>
<t>Based on 15 years of experience and new research, this document
replaces the recommendations of RFC 2309.</t>
</abstract>
</front>
<seriesInfo name="BCP" value="197"/>
<seriesInfo name="RFC" value="7567"/>
<seriesInfo name="DOI" value="10.17487/RFC7567"/>
</reference>
<reference anchor="RFC8033" target="https://www.rfc-editor.org/info/rfc8
033" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8033.xml">
<front>
<title>Proportional Integral Controller Enhanced (PIE): A Lightweigh
t Control Scheme to Address the Bufferbloat Problem</title>
<author initials="R." surname="Pan" fullname="R. Pan">
<organization/>
</author>
<author initials="P." surname="Natarajan" fullname="P. Natarajan">
<organization/>
</author>
<author initials="F." surname="Baker" fullname="F. Baker">
<organization/>
</author>
<author initials="G." surname="White" fullname="G. White">
<organization/>
</author>
<date year="2017" month="February"/>
<abstract>
<t>Bufferbloat is a phenomenon in which excess buffers in the netw
ork cause high latency and latency variation. As more and more interactive appl
ications (e.g., voice over IP, real-time video streaming, and financial transact
ions) run in the Internet, high latency and latency variation degrade applicatio
n performance. There is a pressing need to design intelligent queue management
schemes that can control latency and latency variation, and hence provide desira
ble quality of service to users.</t>
<t>This document presents a lightweight active queue management de
sign called "PIE" (Proportional Integral controller Enhanced) that can effective
ly control the average queuing latency to a target value. Simulation results, th
eoretical analysis, and Linux testbed results have shown that PIE can ensure low
latency and achieve high link utilization under various congestion situations.
The design does not require per-packet timestamps, so it incurs very little ove
rhead and is simple enough to implement in both hardware and software.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="8033"/>
<seriesInfo name="DOI" value="10.17487/RFC8033"/>
</reference>
<reference anchor="RFC8034" target="https://www.rfc-editor.org/info/rfc8
034" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8034.xml">
<front>
<title>Active Queue Management (AQM) Based on Proportional Integral
Controller Enhanced PIE) for Data-Over-Cable Service Interface Specifications (D
OCSIS) Cable Modems</title>
<author initials="G." surname="White" fullname="G. White">
<organization/>
</author>
<author initials="R." surname="Pan" fullname="R. Pan">
<organization/>
</author>
<date year="2017" month="February"/>
<abstract>
<t>Cable modems based on Data-Over-Cable Service Interface Specifi
cations (DOCSIS) provide broadband Internet access to over one hundred million u
sers worldwide. In some cases, the cable modem connection is the bottleneck (lo
west speed) link between the customer and the Internet. As a result, the impact
of buffering and bufferbloat in the cable modem can have a significant effect o
n user experience. The CableLabs DOCSIS 3.1 specification introduces requiremen
ts for cable modems to support an Active Queue Management (AQM) algorithm that i
s intended to alleviate the impact that buffering has on latency-sensitive traff
ic, while preserving bulk throughput performance. In addition, the CableLabs DO
CSIS 3.0 specifications have also been amended to contain similar requirements.
This document describes the requirements on AQM that apply to DOCSIS equipment,
including a description of the "DOCSIS-PIE" algorithm that is required on DOCSI
S 3.1 cable modems.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="8034"/>
<seriesInfo name="DOI" value="10.17487/RFC8034"/>
</reference>
<reference anchor="RFC8174" target="https://www.rfc-editor.org/info/rfc8
174" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8174.xml">
<front>
<title>Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words</ti
tle>
<author initials="B." surname="Leiba" fullname="B. Leiba">
<organization/>
</author>
<date year="2017" month="May"/>
<abstract>
<t>RFC 2119 specifies common key words that may be used in protoco
l specifications. This document aims to reduce the ambiguity by clarifying tha
t only UPPERCASE usage of the key words have the defined special meanings.</t>
</abstract>
</front>
<seriesInfo name="BCP" value="14"/>
<seriesInfo name="RFC" value="8174"/>
<seriesInfo name="DOI" value="10.17487/RFC8174"/>
</reference>
<reference anchor="RFC8257" target="https://www.rfc-editor.org/info/rfc8
257" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8257.xml">
<front>
<title>Data Center TCP (DCTCP): TCP Congestion Control for Data Cent
ers</title>
<author initials="S." surname="Bensley" fullname="S. Bensley">
<organization/>
</author>
<author initials="D." surname="Thaler" fullname="D. Thaler">
<organization/>
</author>
<author initials="P." surname="Balasubramanian" fullname="P. Balasub
ramanian">
<organization/>
</author>
<author initials="L." surname="Eggert" fullname="L. Eggert">
<organization/>
</author>
<author initials="G." surname="Judd" fullname="G. Judd">
<organization/>
</author>
<date year="2017" month="October"/>
<abstract>
<t>This Informational RFC describes Data Center TCP (DCTCP): a TCP
congestion control scheme for data-center traffic. DCTCP extends the Explicit
Congestion Notification (ECN) processing to estimate the fraction of bytes that
encounter congestion rather than simply detecting that some congestion has occur
red. DCTCP then scales the TCP congestion window based on this estimate. This
method achieves high-burst tolerance, low latency, and high throughput with shal
low- buffered switches. This memo also discusses deployment issues related to t
he coexistence of DCTCP and conventional TCP, discusses the lack of a negotiatin
g mechanism between sender and receiver, and presents some possible mitigations.
This memo documents DCTCP as currently implemented by several major operating
systems. DCTCP, as described in this specification, is applicable to deployment
s in controlled environments like data centers, but it must not be deployed over
the public Internet without additional measures.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="8257"/>
<seriesInfo name="DOI" value="10.17487/RFC8257"/>
</reference>
<reference anchor="RFC8298" target="https://www.rfc-editor.org/info/rfc8
298" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8298.xml">
<front>
<title>Self-Clocked Rate Adaptation for Multimedia</title>
<author initials="I." surname="Johansson" fullname="I. Johansson">
<organization/>
</author>
<author initials="Z." surname="Sarker" fullname="Z. Sarker">
<organization/>
</author>
<date year="2017" month="December"/>
<abstract>
<t>This memo describes a rate adaptation algorithm for conversatio
nal media services such as interactive video. The solution conforms to the pack
et conservation principle and uses a hybrid loss-and-delay- based congestion con
trol algorithm. The algorithm is evaluated over both simulated Internet bottlen
eck scenarios as well as in a Long Term Evolution (LTE) system simulator and is
shown to achieve both low latency and high video throughput in these scenarios.<
/t>
</abstract>
</front>
<seriesInfo name="RFC" value="8298"/>
<seriesInfo name="DOI" value="10.17487/RFC8298"/>
</reference>
<reference anchor="RFC8290" target="https://www.rfc-editor.org/info/rfc8
290" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8290.xml">
<front>
<title>The Flow Queue CoDel Packet Scheduler and Active Queue Manage
ment Algorithm</title>
<author initials="T." surname="Hoeiland-Joergensen" fullname="T. Hoe
iland-Joergensen">
<organization/>
</author>
<author initials="P." surname="McKenney" fullname="P. McKenney">
<organization/>
</author>
<author initials="D." surname="Taht" fullname="D. Taht">
<organization/>
</author>
<author initials="J." surname="Gettys" fullname="J. Gettys">
<organization/>
</author>
<author initials="E." surname="Dumazet" fullname="E. Dumazet">
<organization/>
</author>
<date year="2018" month="January"/>
<abstract>
<t>This memo presents the FQ-CoDel hybrid packet scheduler and Act
ive Queue Management (AQM) algorithm, a powerful tool for fighting bufferbloat a
nd reducing latency.</t>
<t>FQ-CoDel mixes packets from multiple flows and reduces the impa
ct of head-of-line blocking from bursty traffic. It provides isolation for low-
rate traffic such as DNS, web, and videoconferencing traffic. It improves utili
sation across the networking fabric, especially for bidirectional traffic, by ke
eping queue lengths short, and it can be implemented in a memory- and CPU-effici
ent fashion across a wide range of hardware.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="8290"/>
<seriesInfo name="DOI" value="10.17487/RFC8290"/>
</reference>
<reference anchor="RFC8312" target="https://www.rfc-editor.org/info/rfc8
312" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8312.xml">
<front>
<title>CUBIC for Fast Long-Distance Networks</title>
<author initials="I." surname="Rhee" fullname="I. Rhee">
<organization/>
</author>
<author initials="L." surname="Xu" fullname="L. Xu">
<organization/>
</author>
<author initials="S." surname="Ha" fullname="S. Ha">
<organization/>
</author>
<author initials="A." surname="Zimmermann" fullname="A. Zimmermann">
<organization/>
</author>
<author initials="L." surname="Eggert" fullname="L. Eggert">
<organization/>
</author>
<author initials="R." surname="Scheffenegger" fullname="R. Scheffene
gger">
<organization/>
</author>
<date year="2018" month="February"/>
<abstract>
<t>CUBIC is an extension to the current TCP standards. It differs
from the current TCP standards only in the congestion control algorithm on the
sender side. In particular, it uses a cubic function instead of a linear window
increase function of the current TCP standards to improve scalability and stabi
lity under fast and long-distance networks. CUBIC and its predecessor algorithm
have been adopted as defaults by Linux and have been used for many years. This
document provides a specification of CUBIC to enable third-party implementation
s and to solicit community feedback through experimentation on the performance o
f CUBIC.</t>
</abstract>
</front>
<seriesInfo name="RFC" value="8312"/>
<seriesInfo name="DOI" value="10.17487/RFC8312"/>
</reference>
<reference anchor="RFC8404" target="https://www.rfc-editor.org/info/rfc8
404" xml:base="https://bib.ietf.org/public/rfc/bibxml/reference.RFC.8404.xml">
<front> <front>
<title>Effects of Pervasive Encryption on Operators</title> <title>Random Early Detection Gateways for Congestion Avoidance</tit
<author initials="K." surname="Moriarty" fullname="K. Moriarty" role le>
="editor"> <author fullname="Sally Floyd" initials="S." surname="Floyd">
<organization/> <organization>UC Berkeley</organization>
</author> </author>
<author initials="A." surname="Morton" fullname="A. Morton" role="ed <author fullname="Van Jacobson" initials="V." surname="Jacobson">
itor"> <organization>UC Berkeley</organization>
<organization/>
</author> </author>
<date year="2018" month="July"/> <date month="August" year="1993"/>
<abstract>
<t>Pervasive monitoring attacks on the privacy of Internet users a
re of serious concern to both user and operator communities. RFC 7258 discusses
the critical need to protect users' privacy when developing IETF specifications
and also recognizes that making networks unmanageable to mitigate pervasive mon
itoring is not an acceptable outcome: an appropriate balance is needed. This do
cument discusses current security and network operations as well as management p
ractices that may be impacted by the shift to increased use of encryption to hel
p guide protocol development in support of manageable and secure networks.</t>
</abstract>
</front> </front>
<seriesInfo name="RFC" value="8404"/> <seriesInfo name="DOI" value="10.1109/90.251892"/>
<seriesInfo name="DOI" value="10.17487/RFC8404"/> <refcontent>IEEE/ACM Transactions on Networking, Volume 1, Issue 4, pp
. 397-413</refcontent>
</reference> </reference>
<reference anchor="ARED01" target="https://www.icir.org/floyd/red.html">
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.2914.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.3246.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.3649.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.5033.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.5348.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.5681.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.5706.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.7567.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8033.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8034.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8174.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8257.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8298.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8290.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8312.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.8404.
xml"/>
<xi:include href="https://xml2rfc.ietf.org/public/rfc/bibxml/reference.RFC.9000.
xml"/>
<reference anchor="ARED01" target="https://www.icsi.berkeley.edu/icsi/no
de/2032">
<front> <front>
<title>Adaptive RED: An Algorithm for Increasing the Robustness of <title>Adaptive RED: An Algorithm for Increasing the Robustness of
RED's Active Queue Management</title> RED's Active Queue Management</title>
<author fullname="Sally Floyd" initials="S." surname="Floyd"> <author fullname="Sally Floyd" initials="S." surname="Floyd">
<organization>ACIRI</organization> <organization>ACIRI</organization>
</author> </author>
<author fullname="Ramakrishna Gummadi" initials="R." surname="Gummad i"> <author fullname="Ramakrishna Gummadi" initials="R." surname="Gummad i">
<organization>ACIRI</organization> <organization>ACIRI</organization>
</author> </author>
<author fullname="S. Shenker" initials="S." surname="Shenker"> <author fullname="S. Shenker" initials="S." surname="Shenker">
<organization>ACIRI</organization> <organization>ACIRI</organization>
</author> </author>
<date month="August" year="2001"/> <date month="August" year="2001"/>
</front> </front>
<seriesInfo name="ACIRI Technical Report" value=""/> <refcontent>ACIRI Technical Report 301</refcontent>
</reference>
<reference anchor="I-D.ietf-tsvwg-l4s-arch" target="https://datatracker.
ietf.org/api/v1/doc/document/draft-ietf-tsvwg-l4s-arch/" xml:base="https://bib.i
etf.org/public/rfc/bibxml-ids/reference.I-D.ietf-tsvwg-l4s-arch.xml">
<front>
<title>Low Latency, Low Loss, Scalable Throughput (L4S) Internet Ser
vice: Architecture</title>
<author fullname="Bob Briscoe"/>
<author fullname="Koen De Schepper"/>
<author fullname="Marcelo Bagnulo"/>
<author fullname="Greg White"/>
<date day="27" month="July" year="2022"/>
<abstract>
<t>This document describes the L4S architecture, which enables Int
ernet
applications to achieve Low queuing Latency, Low Loss, and Scalable
throughput (L4S). The insight on which L4S is based is that the root
cause of queuing delay is in the congestion controllers of senders,
not in the queue itself. With the L4S architecture all Internet
applications could (but do not have to) transition away from
congestion control algorithms that cause substantial queuing delay,
to a new class of congestion controls that induce very little
queuing, aided by explicit congestion signalling from the network.
This new class of congestion controls can provide low latency for
capacity-seeking flows, so applications can achieve both high
bandwidth and low latency.</t>
<t>The architecture primarily concerns incremental deployment. It
defines mechanisms that allow the new class of L4S congestion
controls to coexist with 'Classic' congestion controls in a shared
network. These mechanisms aim to ensure that the latency and
throughput performance using an L4S-compliant congestion controller
is usually much better (and rarely worse) than performance would have
been using a 'Classic' congestion controller, and that competing
flows continuing to use 'Classic' controllers are typically not
impacted by the presence of L4S. These characteristics are important
to encourage adoption of L4S congestion control algorithms and L4S
compliant network elements.</t>
<t>The L4S architecture consists of three components: network supp
ort to
isolate L4S traffic from classic traffic; protocol features that
allow network elements to identify L4S traffic; and host support for
L4S congestion controls. The protocol is defined separately as an
experimental change to Explicit Congestion Notification (ECN).</t>
</abstract>
</front>
<seriesInfo name="Internet-Draft" value="draft-ietf-tsvwg-l4s-arch-19"
/>
</reference>
<reference anchor="I-D.briscoe-tsvwg-l4s-diffserv" target="https://datat
racker.ietf.org/api/v1/doc/document/draft-briscoe-tsvwg-l4s-diffserv/" xml:base=
"https://bib.ietf.org/public/rfc/bibxml-ids/reference.I-D.briscoe-tsvwg-l4s-diff
serv.xml">
<front>
<title>Interactions between Low Latency, Low Loss, Scalable Throughp
ut (L4S) and Differentiated Services</title>
<author fullname="Bob Briscoe"/>
<date day="2" month="July" year="2018"/>
<abstract>
<t>L4S and Diffserv offer somewhat overlapping services (low laten
cy and
low loss), but bandwidth allocation is out of scope for L4S.
Therefore there is scope for the two approaches to complement each
other, but also to conflict. This informational document explains
how the two approaches interact, how they can be arranged to
complement each other and in which cases one can stand alone without
needing the other.</t>
</abstract>
</front>
<seriesInfo name="Internet-Draft" value="draft-briscoe-tsvwg-l4s-diffs
erv-02"/>
</reference>
<reference anchor="I-D.briscoe-docsis-q-protection" target="https://data
tracker.ietf.org/api/v1/doc/document/draft-briscoe-docsis-q-protection/" xml:bas
e="https://bib.ietf.org/public/rfc/bibxml-ids/reference.I-D.briscoe-docsis-q-pro
tection.xml">
<front>
<title>The DOCSIS(r) Queue Protection Algorithm to Preserve Low Late
ncy</title>
<author fullname="Bob Briscoe"/>
<author fullname="Greg White"/>
<date day="13" month="May" year="2022"/>
<abstract>
<t>This informational document explains the specification of the q
ueue
protection algorithm used in DOCSIS technology since version 3.1. A
shared low latency queue relies on the non-queue-building behaviour
of every traffic flow using it. However, some flows might not take
such care, either accidentally or maliciously. If a queue is about
to exceed a threshold level of delay, the queue protection algorithm
can rapidly detect the flows most likely to be responsible. It can
then prevent harm to other traffic in the low latency queue by
ejecting selected packets (or all packets) of these flows. The
document is designed for four types of audience: a) congestion
control designers who need to understand how to keep on the 'good'
side of the algorithm; b) implementers of the algorithm who want to
understand it in more depth; c) designers of algorithms with similar
goals, perhaps for non-DOCSIS scenarios; and d) researchers
interested in evaluating the algorithm.</t>
</abstract>
</front>
<seriesInfo name="Internet-Draft" value="draft-briscoe-docsis-q-protec
tion-06"/>
</reference>
<reference anchor="I-D.cardwell-iccrg-bbr-congestion-control" target="ht
tps://datatracker.ietf.org/api/v1/doc/document/draft-cardwell-iccrg-bbr-congesti
on-control/" xml:base="https://bib.ietf.org/public/rfc/bibxml-ids/reference.I-D.
cardwell-iccrg-bbr-congestion-control.xml">
<front>
<title>BBR Congestion Control</title>
<author fullname="Neal Cardwell"/>
<author fullname="Yuchung Cheng"/>
<author fullname="Soheil Hassas Yeganeh"/>
<author fullname="Ian Swett"/>
<author fullname="Van Jacobson"/>
<date day="7" month="March" year="2022"/>
<abstract>
<t>This document specifies the BBR congestion control algorithm.
BBR
("Bottleneck Bandwidth and Round-trip propagation time") uses recent
measurements of a transport connection's delivery rate, round-trip
time, and packet loss rate to build an explicit model of the network
path. BBR then uses this model to control both how fast it sends
data and the maximum volume of data it allows in flight in the
network at any time. Relative to loss-based congestion control
algorithms such as Reno [RFC5681] or CUBIC [RFC8312], BBR offers
substantially higher throughput for bottlenecks with shallow buffers
or random losses, and substantially lower queueing delays for
bottlenecks with deep buffers (avoiding "bufferbloat"). BBR can be
implemented in any transport protocol that supports packet-delivery
acknowledgment. Thus far, open source implementations are available
for TCP [RFC793] and QUIC [RFC9000]. This document specifies version
2 of the BBR algorithm, also sometimes referred to as BBRv2 or bbr2.</t>
</abstract>
</front>
<seriesInfo name="Internet-Draft" value="draft-cardwell-iccrg-bbr-cong
estion-control-02"/>
</reference>
<reference anchor="I-D.briscoe-iccrg-prague-congestion-control" target="
https://datatracker.ietf.org/api/v1/doc/document/draft-briscoe-iccrg-prague-cong
estion-control/" xml:base="https://bib.ietf.org/public/rfc/bibxml-ids/reference.
I-D.briscoe-iccrg-prague-congestion-control.xml">
<front>
<title>Prague Congestion Control</title>
<author fullname="Koen De Schepper"/>
<author fullname="Olivier Tilmans"/>
<author fullname="Bob Briscoe"/>
<date day="11" month="July" year="2022"/>
<abstract>
<t>This specification defines the Prague congestion control scheme
,
which is derived from DCTCP and adapted for Internet traffic by
implementing the Prague L4S requirements. Over paths with L4S
support at the bottleneck, it adapts the DCTCP mechanisms to achieve
consistently low latency and full throughput. It is defined
independently of any particular transport protocol or operating
system, but notes are added that highlight issues specific to certain
transports and OSs. It is mainly based on the current default
options of the reference Linux implementation of TCP Prague, but it
includes experience from other implementations where available. It
separately describes non-default and optional parts, as well as
future plans.</t>
<t>The implementation does not satisfy all the Prague requirements
(yet)
and the IETF might decide that certain requirements need to be
relaxed as an outcome of the process of trying to satisfy them all.
In two cases, research code is replaced by placeholders until full
evaluation is complete.</t>
</abstract>
</front>
<seriesInfo name="Internet-Draft" value="draft-briscoe-iccrg-prague-co
ngestion-control-01"/>
</reference>
<reference anchor="I-D.mathis-iccrg-relentless-tcp" target="https://www.
ietf.org/archive/id/draft-mathis-iccrg-relentless-tcp-00.txt" xml:base="https://
bib.ietf.org/public/rfc/bibxml-ids/reference.I-D.mathis-iccrg-relentless-tcp.xml
">
<front>
<title>Relentless Congestion Control</title>
<author fullname="Matt Mathis"/>
<date day="4" month="March" year="2009"/>
<abstract>
<t>Relentless congestion control is a simple modification that can
be applied to almost any AIMD style congestion control: instead of applying a m
ultiplicative reduction to cwnd after a loss, cwnd is reduced by the number of l
ost segments. It can be modeled as a strict implementation of van Jacobson's Pac
ket Conservation Principle. During recovery, new segments are injected into the
network in exact accordance with the segments that are reported to have been del
ivered to the receiver by the returning ACKs. This algorithm offers a valuable n
ew congestion control property: the TCP portion of the control loop has exactly
unity gain, which should make it easier to implement simple controllers in netwo
rk devices to accurately control queue sizes across a huge range of scales. Rele
ntless Congestion Control conforms to neither the details nor the philosophy of
current congestion control standards. These standards are based on the idea that
the Internet can attain sufficient fairness by having relatively simple network
devices send uniform congestion signals to all flows, and mandating that all pr
otocols have equivalent responses to these congestion signals. To function appro
priately in a shared environment, Relentless Congestion Control requires that th
e network allocates capacity through some technique such as Fair Queuing, Approx
imate Fair Dropping, etc. The salient features of these algorithms are that they
segregate the traffic into distinct flows, and send different congestion signal
s to each flow. This alternative congestion control paradigm is described in a s
eparate document, also under consideration by the ICCRG. The goal of the documen
t is to illustrate some new protocol features and properties might be possible i
f we relax the "TCP-friendly" mandate. A secondary goal of Relentless TCP is to
make a distinction between the bottlenecks that belong to protocol itself, vs st
andard congestion control and the "TCP-friendly" paradigm.</t>
</abstract>
</front>
<seriesInfo name="Internet-Draft" value="draft-mathis-iccrg-relentless
-tcp-00"/>
</reference> </reference>
<!--{ToDo: DCttH ref will need to be updated, once stable}-->
<reference anchor="DCttH19" target="https://bobbriscoe.net/pubs.html#DCttH <!-- [I-D.ietf-tsvwg-l4s-arch] companion doc 9330 - title matches as of 1/17/23-
_TR"> ->
<front> <reference anchor='RFC9330' target='https://www.rfc-editor.org/info/rfc9330'>
<title>`Data Centre to the Home': Ultra-Low Latency for All</title> <front>
<author fullname="Koen De Schepper" initials="K." surname="De Schepp <title>Low Latency, Low Loss, and Scalable Throughput (L4S) Internet Service: Ar
er"> chitecture</title>
<organization>Nokia Bell Labs</organization> <author initials='B' surname='Briscoe' fullname='Bob Briscoe' role='editor'>
</author> </author>
<author fullname="Olga Bondarenko" initials="O." surname="Bondarenko <author initials='K' surname='De Schepper' fullname='Koen De Schepper'>
"> </author>
<organization>Simula Research Lab</organization> <author initials='M' surname='Bagnulo' fullname='Marcelo Bagnulo'>
</author> </author>
<author fullname="Olivier" initials="O." surname="Tilmans"> <author initials='G' surname='White' fullname='Greg White'>
<organization>Nokia Bell Labs</organization> </author>
</author> <date year='2023' month='January'/>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> </front>
<organization>Independent (bobbriscoe.net)</organization> <seriesInfo name="RFC" value="9330"/>
</author> <seriesInfo name="DOI" value="10.17487/RFC9330"/>
<date month="July" year="2019"/> </reference>
</front>
<seriesInfo name="Updated RITE project Technical Report" value=""/> <!-- [I-D.briscoe-tsvwg-l4s-diffserv] IESG state Expired as of 1/17/23 -->
<format target="https://bobbriscoe.net/projects/latency/dctth_journal_ <xi:include href="https://datatracker.ietf.org/doc/bibxml3/reference.I-D.briscoe
draft20190726.pdf" type="PDF"/> -tsvwg-l4s-diffserv.xml"/>
</reference>
<reference anchor="PI2" target="https://riteproject.files.wordpress.com/ <!-- [I-D.briscoe-docsis-q-protection] in MISSREF state as of 1/17/23 -->
2015/10/pi2_conext.pdf"> <xi:include href="https://datatracker.ietf.org/doc/bibxml3/reference.I-D.briscoe
-docsis-q-protection.xml"/>
<!-- [I-D.cardwell-iccrg-bbr-congestion-control] IESG state Expired as of 1/17/2
3 -->
<xi:include href="https://datatracker.ietf.org/doc/bibxml3/reference.I-D.cardwel
l-iccrg-bbr-congestion-control.xml"/>
<!-- [I-D.briscoe-iccrg-prague-congestion-control] IESG state Expired as of 1/17
/23 -->
<xi:include href="https://datatracker.ietf.org/doc/bibxml3/reference.I-D.briscoe
-iccrg-prague-congestion-control.xml"/>
<!-- [I-D.mathis-iccrg-relentless-tcp] IESG state Expired as of 1/17/23 -->
<xi:include href="https://datatracker.ietf.org/doc/bibxml3/reference.I-D.mathis-
iccrg-relentless-tcp.xml"/>
<reference anchor="PI2" target="https://dl.acm.org/doi/10.1145/2999572.2
999578">
<front> <front>
<title>PI2: A Linearized AQM for both Classic and Scalable <title>PI2: A Linearized AQM for both Classic and Scalable
TCP</title> TCP</title>
<author fullname="Koen De Schepper" initials="K." surname="De Schepp er"> <author fullname="Koen De Schepper" initials="K." surname="De Schepp er">
<organization>Nokia Bell Labs</organization> <organization>Nokia Bell Labs</organization>
</author> </author>
<author fullname="Olga Bondarenko" initials="O." surname="Bondarenko "> <author fullname="Olga Bondarenko" initials="O." surname="Bondarenko ">
<organization>Simula Research Lab</organization> <organization>Simula Research Lab</organization>
</author> </author>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> <author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization>BT</organization> <organization>BT</organization>
</author> </author>
<author fullname="Ing-jyh Tsang" initials="I." surname="Tsang"> <author fullname="Ing-jyh Tsang" initials="I." surname="Tsang">
<organization>Nokia Bell Labs</organization> <organization>Nokia Bell Labs</organization>
</author> </author>
<date month="December" year="2016"/> <date month="December" year="2016"/>
</front> </front>
<seriesInfo name="ACM CoNEXT'16" value=""/> <seriesInfo name="DOI" value="10.1145/2999572.2999578"/>
<refcontent>ACM CoNEXT'16</refcontent>
</reference> </reference>
<reference anchor="L4Sdemo16" target="https//dl.acm.org/citation.cfm?doi
d=2910017.2910633 (videos of demos: https://riteproject.eu/dctth/#1511dispatchwg <reference anchor="L4Sdemo16" target="https://dl.acm.org/citation.cfm?do
)"> id=2910017.2910633">
<front> <front>
<title>Ultra-Low Delay for All: Live Experience, Live <title>Ultra-Low Delay for All: Live Experience, Live Analysis</titl
Analysis</title> e>
<author fullname="Olga Bondarenko" initials="O." surname="Bondarenko "> <author fullname="Olga Bondarenko" initials="O." surname="Bondarenko ">
<organization>Simula Research Lab</organization> <organization>Simula Research Lab</organization>
</author> </author>
<author fullname="Koen De Schepper" initials="K." surname="De Schepp er"> <author fullname="Koen De Schepper" initials="K." surname="De Schepp er">
<organization>Bell Labs</organization> <organization>Bell Labs</organization>
</author> </author>
<author fullname="Ing-jyh Tsang" initials="I." surname="Tsang"> <author fullname="Ing-jyh Tsang" initials="I." surname="Tsang">
<organization>Bell Labs</organization> <organization>Bell Labs</organization>
</author> </author>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> <author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization>BT</organization> <organization>BT</organization>
</author> </author>
<date month="May" year="2016"/> <author fullname="Andreas Petlund" initials="A." surname="Petlund">
</author>
<author fullname="Carsten Griwodz" initials="C." surname="Griwodz">
</author>
<date month="May" year="2016"/>
</front> </front>
<seriesInfo name="Proc. MMSYS'16" value="pp33:1--33:4"/> <seriesInfo name="DOI" value="10.1145/2910017.2910633"/>
<refcontent>Proceedings of the 7th International Conference on Multime
dia Systems, Article No. 33, pp. 1-4</refcontent>
<format target="https://dl.acm.org/citation.cfm?doid=2910017.2910633" type="PDF"/> <format target="https://dl.acm.org/citation.cfm?doid=2910017.2910633" type="PDF"/>
</reference> </reference>
<reference anchor="Dukkipati06" target="https://dl.acm.org/doi/10.1145/1
111322.1111336"> <reference anchor="L4Seval22" target="https://arxiv.org/abs/2209.01078">
<front>
<title>Dual Queue Coupled AQM: Deployable Very Low Queuing Delay for
All</title>
<author fullname="Koen De Schepper" initials="K." surname="De Schepper
">
<organization>Nokia Bell Labs</organization>
</author>
<author fullname="Olga Albisser" initials="O." surname="Albisser">
<organization>Simula Research Lab</organization>
</author>
<author fullname="Olivier Tilmans" initials="O." surname="Tilmans">
<organization>Nokia Bell Labs</organization>
</author>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization>Independent (bobbriscoe.net)</organization>
</author>
<date month="September" year="2022"/>
</front>
<seriesInfo name="DOI" value="10.48550/arXiv.2209.01078"/>
<refcontent>Preprint submitted to IEEE/ACM Transactions on Networking</r
efcontent>
</reference>
<reference anchor="Dukkipati06" target="https://dl.acm.org/doi/10.1145/1111322.1
111336">
<front> <front>
<title>Why Flow-Completion Time is the Right Metric for Congestion <title>Why Flow-Completion Time is the Right Metric for Congestion C
Control</title> ontrol</title>
<author fullname="Nandita Dukkipati" initials="N." surname="Dukkipat i"> <author fullname="Nandita Dukkipati" initials="N." surname="Dukkipat i">
<organization>Stanford Uni</organization> <organization>Stanford University</organization>
</author> </author>
<author fullname="Nick McKeown" initials="N." surname="McKeown"> <author fullname="Nick McKeown" initials="N." surname="McKeown">
<organization>Stanford Uni</organization> <organization>Stanford University</organization>
</author> </author>
<date month="January" year="2006"/> <date month="January" year="2006"/>
</front> </front>
<seriesInfo name="ACM CCR" value="36(1):59--62"/> <seriesInfo name="DOI" value="10.1145/1111322.1111336"/>
<format target="http://yuba.stanford.edu/rcp/flowCompTime-dukkipati.pd <refcontent>ACM SIGCOMM Computer Communication Review, Vol. 36, Issue
f" type="PDF"/> 1, pp. 59-62</refcontent>
</reference> </reference>
<!-- <reference anchor="DCTCP_Pitfalls"
target="http://blogs.usenix.org/conference/nsdi15/technical-ses
sions/presentation/judd">
<front>
<title>Attaining the Promise and Avoiding the Pitfalls of TCP in the
Datacenter</title>
<author fullname="Glenn Judd" initials="G." surname="Judd">
<organization>Morgan Stanley</organization>
</author>
<date month="May" year="2015"/>
</front>
<seriesInfo name="12th USENIX Symposium on Networked Systems Design and
Implementation (NSDI 15)"
value="145-157"/>
<format target="http://blogs.usenix.org/conference/nsdi15/technical-sess
ions/presentation/judd"
type="PDF"/>
</reference>
<reference anchor="CRED_Insights" target="https://arxiv.org/abs/1904.07339 "> <reference anchor="CRED_Insights" target="https://arxiv.org/abs/1904.07339 ">
<front> <front>
<title>Insights from Curvy RED (Random Early Detection)</title> <title>Insights from Curvy RED (Random Early Detection)</title>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> <author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization>BT</organization> <organization>BT</organization>
</author> </author>
<date day="" month="July" year="2015"/> <author fullname="Koen De Schepper" initials="K." surname="De Scheppe
r">
<organization>BT</organization>
</author>
<date month="August" year="2015"/>
</front> </front>
<seriesInfo name="BT Technical Report" value="TR-TUB8-2015-003 arXiv:1 <seriesInfo name="DOI" value="10.48550/arXiv.1904.07339"/>
904.07339 [cs.NI]"/> <refcontent>BT Technical Report, TR-TUB8-2015-003</refcontent>
<format target="https://arxiv.org/pdf/1904.07339" type="PDF"/> <format target="https://arxiv.org/pdf/1904.07339" type="PDF"/>
</reference> </reference>
<reference anchor="CoDel" target="https://queue.acm.org/issuedetail.cfm? issue=2208917"> <reference anchor="CoDel" target="https://queue.acm.org/issuedetail.cfm? issue=2208917">
<front> <front>
<title>Controlling Queue Delay</title> <title>Controlling Queue Delay</title>
<author fullname="Kathleen Nichols" initials="K." surname="Nichols"> <author fullname="Kathleen Nichols" initials="K." surname="Nichols">
<organization>PARC</organization> <organization>PARC</organization>
</author> </author>
<author fullname="Van Jacobson" initials="V." surname="Jacobson"> <author fullname="Van Jacobson" initials="V." surname="Jacobson">
<organization>Pollere Inc</organization> <organization>Pollere Inc</organization>
</author> </author>
<date month="May" year="2012"/> <date month="May" year="2012"/>
</front> </front>
<seriesInfo name="ACM Queue" value="10(5)"/> <refcontent>ACM Queue, Vol. 10, Issue 5</refcontent>
</reference> </reference>
<reference anchor="MEDF" target="https://infocom2003.ieee-infocom.org/pa
pers/27_04.PDF"> <reference anchor="MEDF">
<front> <front>
<title>MEDF - a simple scheduling algorithm for two real-time <title>MEDF - A Simple Scheduling Algorithm for Two Real-Time
transport service classes with application in the UTRAN</title> Transport Service Classes with Application in the UTRAN</title>
<author fullname="Michael Menth " initials="M." surname="Menth"> <author fullname="Michael Menth " initials="M." surname="Menth">
<organization>University of Wuerzburg</organization> <organization>University of Wuerzburg</organization>
</author> </author>
<author fullname="Matthias Schmid " initials="M." surname="Schmid"> <author fullname="Matthias Schmid " initials="M." surname="Schmid">
<organization>Infosim AG</organization> <organization>Infosim AG</organization>
</author> </author>
<author fullname="Herbert Heiss" initials="H." surname="Heiss"> <author fullname="Herbert Heiss" initials="H." surname="Heiss">
<organization>Siemens</organization> <organization>Siemens</organization>
</author> </author>
<author fullname="Thomas Reim" initials="T." surname="Reim"> <author fullname="Thomas Reim" initials="T." surname="Reim">
<organization>Siemens</organization> <organization>Siemens</organization>
</author> </author>
<date month="March" year="2003"/> <date month="March" year="2003"/>
</front> </front>
<seriesInfo name="Proc. IEEE Conference on Computer Communications (IN <seriesInfo name="DOI" value="10.1109/INFCOM.2003.1208948"/>
FOCOM'03)" value="Vol.2 pp.1116-1122"/> <refcontent>Proc. IEEE Conference on Computer Communications (INFOCOM'
03), Vol. 2, pp. 1116-1122</refcontent>
</reference> </reference>
<reference anchor="DualQ-Test" target="https://www.duo.uio.no/bitstream/
handle/10852/57424/thesis-henrste.pdf?sequence=1"> <reference anchor="DualQ-Test">
<front> <front>
<title>Destruction Testing: Ultra-Low Delay using Dual Queue Coupled <title>Destruction Testing: Ultra-Low Delay using Dual Queue Coupled
Active Queue Management</title> Active Queue Management</title>
<author fullname="Henrik Steen" initials="H." surname="Steen"> <author fullname="Henrik Steen" initials="H." surname="Steen">
<organization>Uni Oslo</organization> <organization>University of Oslo</organization>
</author> </author>
<date month="May" year="2017"/> <date month="May" year="2017"/>
</front> </front>
<seriesInfo name="Master's Thesis, Dept of Informatics, Uni Oslo" valu e=""/> <refcontent>Master's Thesis, Department of Informatics, University of Oslo</refcontent>
</reference> </reference>
<reference anchor="SigQ-Dyn" target="https://arxiv.org/abs/1904.07044"> <reference anchor="SigQ-Dyn" target="https://arxiv.org/abs/1904.07044">
<front> <front>
<title>Rapid Signalling of Queue Dynamics</title> <title>Rapid Signalling of Queue Dynamics</title>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> <author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization/> <organization/>
</author> </author>
<date month="September" year="2017"/> <date month="September" year="2017"/>
</front> </front>
<seriesInfo name="Technical Report" value="TR-BB-2017-001 arXiv:1904.0 <seriesInfo name="DOI" value="10.48550/arXiv.1904.07044"/>
7044 [cs.NI]"/> <refcontent>Technical Report, TR-BB-2017-001</refcontent>
<format target="https://arxiv.org/pdf/1904.07044" type="PDF"/> <format target="https://arxiv.org/pdf/1904.07044" type="PDF"/>
</reference> </reference>
<reference anchor="Alizadeh-stability" target="https://dl.acm.org/citati on.cfm?id=1993753"> <reference anchor="Alizadeh-stability" target="https://dl.acm.org/citati on.cfm?id=1993753">
<front> <front>
<title>Analysis of DCTCP: Stability, Convergence, and <title>Analysis of DCTCP: Stability, Convergence, and
Fairness</title> Fairness</title>
<author fullname="Mohamed Alizadeh" initials="M." surname="Alizadeh" /> <author fullname="Mohamed Alizadeh" initials="M." surname="Alizadeh" />
<author fullname="Adel Javanmard" initials="A." surname="Javanmard"/ > <author fullname="Adel Javanmard" initials="A." surname="Javanmard"/ >
<author fullname="Balaji Prabhakar" initials="B." surname="Prabhakar "/> <author fullname="Balaji Prabhakar" initials="B." surname="Prabhakar "/>
<date month="June" year="2011"/> <date month="June" year="2011"/>
</front> </front>
<seriesInfo name="ACM SIGMETRICS 2011" value=""/> <seriesInfo name="DOI" value="10.1145/1993744.1993753"/>
<refcontent>SIGMETRICS '11: Proceedings of the ACM SIGMETRICS Joint In
ternational Conference on Measurement and Modeling of Computer Systems, pp. 73-8
4</refcontent>
<format target="https://people.csail.mit.edu/alizadeh/papers/dctcp_ana lysis-sigmetrics11.pdf" type="PDF"/> <format target="https://people.csail.mit.edu/alizadeh/papers/dctcp_ana lysis-sigmetrics11.pdf" type="PDF"/>
</reference> </reference>
<reference anchor="PragueLinux" target="https://www.netdevconf.org/0x13/ session.html?talk-tcp-prague-l4s"> <reference anchor="PragueLinux" target="https://www.netdevconf.org/0x13/ session.html?talk-tcp-prague-l4s">
<front> <front>
<title>Implementing the `TCP Prague' Requirements for Low Latency <title>Implementing the 'TCP Prague' Requirements for L4S</title>
Low Loss Scalable Throughput (L4S)</title>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> <author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization>Independent</organization> <organization>Independent</organization>
</author> </author>
<author fullname="Koen De Schepper" initials="K." surname="De Schepp er"> <author fullname="Koen De Schepper" initials="K." surname="De Schepp er">
<organization>Nokia Bell Labs</organization> <organization>Nokia Bell Labs</organization>
</author> </author>
<author fullname="Olga Albisser" initials="O." surname="Albisser"> <author fullname="Olga Albisser" initials="O." surname="Albisser">
<organization>Simula Research Lab</organization> <organization>Simula Research Lab</organization>
</author> </author>
<author fullname="Joakim Misund" initials="J." surname="Misund"> <author fullname="Joakim Misund" initials="J." surname="Misund">
<organization>Simula Research Lab</organization> <organization>Simula Research Lab</organization>
</author> </author>
<author fullname="Olivier Tilmans" initials="O." surname="Tilmans"> <author fullname="Olivier Tilmans" initials="O." surname="Tilmans">
<organization>Nokia Bell Labs</organization> <organization>Nokia Bell Labs</organization>
</author> </author>
<author fullname="Mirja Kühlewind" initials="M." surname="Kühlewind" > <author fullname="Mirja Kuehlewind" initials="M." surname="Kuehlewin d">
<organization>ETH Zurich</organization> <organization>ETH Zurich</organization>
</author> </author>
<author fullname="Asad Sajjad Ahmed" initials="A.S." surname="Ahmed" > <author fullname="Asad Sajjad Ahmed" initials="A." surname="Ahmed">
<organization>Simula Research Lab</organization> <organization>Simula Research Lab</organization>
</author> </author>
<date month="March" year="2019"/> <date month="March" year="2019"/>
</front> </front>
<seriesInfo name="Proc. Linux Netdev 0x13" value=""/> <refcontent>Proceedings of Linux Netdev 0x13</refcontent>
<format target="https://www.files.netdevconf.org/f/4d6939d5f1fb404fafd
1/?dl=1" type="PDF"/>
</reference> </reference>
<reference anchor="DualPI2Linux" target="https://www.netdevconf.org/0x13 /session.html?talk-DUALPI2-AQM"> <reference anchor="DualPI2Linux" target="https://www.netdevconf.org/0x13 /session.html?talk-DUALPI2-AQM">
<front> <front>
<title>DUALPI2 - Low Latency, Low Loss and Scalable (L4S) <title>DUALPI2 - Low Latency, Low Loss and Scalable (L4S)
AQM</title> AQM</title>
<author fullname="Olga Albisser" initials="O." surname="Albisser"> <author fullname="Olga Albisser" initials="O." surname="Albisser">
<organization>Simula Research Lab</organization> <organization>Simula Research Lab</organization>
</author> </author>
<author fullname="Koen De Schepper" initials="K." surname="De Schepp er"> <author fullname="Koen De Schepper" initials="K." surname="De Schepp er">
<organization>Nokia Bell Labs</organization> <organization>Nokia Bell Labs</organization>
</author> </author>
skipping to change at line 2210 skipping to change at line 1643
<organization>Independent</organization> <organization>Independent</organization>
</author> </author>
<author fullname="Olivier Tilmans" initials="O." surname="Tilmans"> <author fullname="Olivier Tilmans" initials="O." surname="Tilmans">
<organization>Nokia Bell Labs</organization> <organization>Nokia Bell Labs</organization>
</author> </author>
<author fullname="Henrik Steen" initials="H." surname="Steen"> <author fullname="Henrik Steen" initials="H." surname="Steen">
<organization>Simula Research Lab</organization> <organization>Simula Research Lab</organization>
</author> </author>
<date month="March" year="2019"/> <date month="March" year="2019"/>
</front> </front>
<seriesInfo name="Proc. Linux Netdev 0x13" value=""/> <seriesInfo name="Proceedings of Linux Netdev 0x13" value=""/>
<format target="https://www.files.netdevconf.org/f/febbe8c6a05b4ceab64 1/?dl=1" type="PDF"/> <format target="https://www.files.netdevconf.org/f/febbe8c6a05b4ceab64 1/?dl=1" type="PDF"/>
</reference> </reference>
<reference anchor="DOCSIS3.1" target="https://specification-search.cable
labs.com/CM-SP-MULPIv3.1"> <reference anchor="DOCSIS3.1" target="https://specification-search.cable
labs.com/CM-SP-MULPIv3">
<front> <front>
<title>MAC and Upper Layer Protocols Interface (MULPI) <title>DOCSIS 3.1 MAC and Upper Layer Protocols Interface Specificat
Specification, CM-SP-MULPIv3.1</title> ion</title>
<author fullname="" surname=""> <author fullname="" surname="">
<organization>CableLabs</organization> <organization>CableLabs</organization>
</author> </author>
<date day="21" month="January" year="2019"/> <date month="January" year="2019"/>
</front> </front>
<seriesInfo name="Data-Over-Cable Service Interface Specifications DOC SIS® 3.1" value="Version i17 or later"/> <refcontent>CM-SP-MULPIv3.1, Data-Over-Cable Service Interface Specifi cations DOCSIS 3.1 Version I17 or later</refcontent>
</reference> </reference>
<reference anchor="LLD" target="https://cablela.bs/low-latency-docsis-te chnology-overview-february-2019"> <reference anchor="LLD" target="https://cablela.bs/low-latency-docsis-te chnology-overview-february-2019">
<front> <front>
<title>Low Latency DOCSIS: Technology Overview</title> <title>Low Latency DOCSIS: Technology Overview</title>
<author fullname="Greg White" initials="G." surname="White"> <author fullname="Greg White" initials="G." surname="White">
<organization>CableLabs</organization> <organization>CableLabs</organization>
</author> </author>
<author fullname="Karthik Sundaresan" initials="K." surname="Sundare san"> <author fullname="Karthik Sundaresan" initials="K." surname="Sundare san">
<organization>CableLabs</organization> <organization>CableLabs</organization>
</author> </author>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> <author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization>CableLabs</organization> <organization>CableLabs</organization>
</author> </author>
<date day="" month="February" year="2019"/> <date day="" month="February" year="2019"/>
</front> </front>
<seriesInfo name="CableLabs White Paper" value=""/> <refcontent>CableLabs White Paper</refcontent>
</reference> </reference>
<reference anchor="AQMmetrics" target="https://www.cs.purdue.edu/homes/f ahmy/papers/ldc.pdf"> <reference anchor="AQMmetrics" target="https://www.cs.purdue.edu/homes/f ahmy/papers/ldc.pdf">
<front> <front>
<title>A Comparison of Load-based and Queue- based Active Queue <title>A Comparison of Load-based and Queue-based Active Queue
Management Algorithms</title> Management Algorithms</title>
<author fullname="Minseok Kwon" initials="M." surname="Kwon"> <author fullname="Minseok Kwon" initials="M." surname="Kwon">
<organization>Purdue Uni</organization> <organization>Purdue University</organization>
</author> </author>
<author fullname="Sonia Fahmy" initials="S." surname="Fahmy"> <author fullname="Sonia Fahmy" initials="S." surname="Fahmy">
<organization>Purdue Uni</organization> <organization>Purdue University</organization>
</author> </author>
<date year="2002"/> <date year="2002"/>
</front> </front>
<seriesInfo name="Proc. Int'l Soc. for Optical Engineering (SPIE)" val <seriesInfo name="DOI" value="10.1117/12.473021"/>
ue="4866:35--46 DOI: 10.1117/12.473021"/> <refcontent>Proc. Int'l Soc. for Optical Engineering (SPIE), Vol. 4866
, pp. 35-46</refcontent>
</reference> </reference>
<reference anchor="CCcensus19" target="https://doi.org/10.1145/3366693"> <reference anchor="CCcensus19" target="https://doi.org/10.1145/3366693">
<front> <front>
<title>The Great Internet TCP Congestion Control Census</title> <title>The Great Internet TCP Congestion Control Census</title>
<author fullname="Ayush Mishra" initials="A." surname="Mishra"> <author fullname="Ayush Mishra" initials="A." surname="Mishra">
<organization/> <organization/>
</author> </author>
<author fullname="Xiangpeng Sun" initials="X." surname="Sun"> <author fullname="Xiangpeng Sun" initials="X." surname="Sun">
<organization/>
<address>
<postal>
<street/>
<city/>
<region/>
<code/>
<country/>
</postal>
<phone/>
<email/>
<uri/>
</address>
</author> </author>
<author fullname="Atishya Jain" initials="A." surname="Jain"> <author fullname="Atishya Jain" initials="A." surname="Jain">
<organization/>
<address>
<postal>
<street/>
<city/>
<region/>
<code/>
<country/>
</postal>
<phone/>
<email/>
<uri/>
</address>
</author> </author>
<author fullname="Sameer Pande" initials="S." surname="Pande"> <author fullname="Sameer Pande" initials="S." surname="Pande">
<organization/>
<address>
<postal>
<street/>
<city/>
<region/>
<code/>
<country/>
</postal>
<phone/>
<email/>
<uri/>
</address>
</author> </author>
<author fullname="Raj Joshi" initials="R." surname="Joshi"> <author fullname="Raj Joshi" initials="R." surname="Joshi">
<organization/>
<address>
<postal>
<street/>
<city/>
<region/>
<code/>
<country/>
</postal>
<phone/>
<email/>
<uri/>
</address>
</author> </author>
<author fullname="Ben Leong" initials="B." surname="Leong"> <author fullname="Ben Leong" initials="B." surname="Leong">
<organization/>
<address>
<postal>
<street/>
<city/>
<region/>
<code/>
<country/>
</postal>
<phone/>
<email/>
<uri/>
</address>
</author> </author>
<date month="December" year="2019"/> <date month="December" year="2019"/>
</front> </front>
<seriesInfo name="Proc. ACM on Measurement and Analysis of Computing S <seriesInfo name="DOI" value="10.1145/3366693"/>
ystems" value="3(3)"/> <refcontent>Proceedings of the ACM on Measurement and Analysis of Comp
uting Systems, Vol. 3, Issue 3, Article No. 45, pp. 1-24</refcontent>
<format target="https://dl.acm.org/doi/10.1145/3366693" type="PDF"/> <format target="https://dl.acm.org/doi/10.1145/3366693" type="PDF"/>
</reference> </reference>
<reference anchor="PI2param" target="https://arxiv.org/abs/2107.01003"> <reference anchor="PI2param" target="https://arxiv.org/abs/2107.01003">
<front> <front>
<title>PI2 Parameters</title> <title>PI2 Parameters</title>
<author fullname="Bob Briscoe" initials="B." surname="Briscoe"> <author fullname="Bob Briscoe" initials="B." surname="Briscoe">
<organization/> <organization/>
</author> </author>
<date month="July" year="2021"/> <date month="July" year="2021"/>
</front> </front>
<seriesInfo name="Technical Report" value="TR-BB-2021-001 arXiv:2107.0 <seriesInfo name="DOI" value="10.48550/arXiv.2107.01003"/>
1003 [cs.NI]"/> <refcontent>Technical Report, TR-BB-2021-001, arXiv:2107.01003 [cs.NI]
</refcontent>
<format target="https://arxiv.org/pdf/2107.01003" type="PDF"/> <format target="https://arxiv.org/pdf/2107.01003" type="PDF"/>
</reference> </reference>
<reference anchor="Labovitz10" target="https://doi.org/10.1145/1851275.1 851194"> <reference anchor="Labovitz10" target="https://doi.org/10.1145/1851275.1 851194">
<front> <front>
<title>Internet Inter-Domain Traffic</title> <title>Internet Inter-Domain Traffic</title>
<author fullname="Craig Labovitz" initials="C." surname="Labovitz"> <author fullname="Craig Labovitz" initials="C." surname="Labovitz">
<organization>Arbor Networks</organization> <organization>Arbor Networks</organization>
</author> </author>
<author fullname="Scott Iekel-Johnson" initials="S." surname="Iekel- Johnson"> <author fullname="Scott Iekel-Johnson" initials="S." surname="Iekel- Johnson">
<organization>Arbor Networks</organization> <organization>Arbor Networks</organization>
</author> </author>
<author fullname="Danny McPherson" initials="D." surname="McPherson" > <author fullname="Danny McPherson" initials="D." surname="McPherson" >
<organization>Arbor Networks</organization> <organization>Arbor Networks</organization>
</author> </author>
<author fullname="Jon Oberheide" initials="J." surname="Oberheide"> <author fullname="Jon Oberheide" initials="J." surname="Oberheide">
<organization>Uni Michigan</organization> <organization>Uni Michigan</organization>
</author> </author>
<author fullname="Farnam Jahanian" initials="F." surname="Jahanian"> <author fullname="Farnam Jahanian" initials="F." surname="Jahanian">
<organization>Uni Michigan</organization> <organization>Uni Michigan</organization>
</author> </author>
<date month="August" year="2010"/> <date month="August" year="2010"/>
</front> </front>
<seriesInfo name="Proc ACM SIGCOMM; ACM CCR" value="40(4):75--86"/> <seriesInfo name="DOI" value="10.1145/1851275.1851194"/>
<refcontent>ACM SIGCOMM Computer Communication Review, Vol. 40, Issue
4, pp. 75-86</refcontent>
<format target="https://dl.acm.org/doi/pdf/10.1145/1851182.1851194" ty pe="PDF"/> <format target="https://dl.acm.org/doi/pdf/10.1145/1851182.1851194" ty pe="PDF"/>
</reference> </reference>
<reference anchor="SCReAM" target="https://github.com/EricssonResearch/s
cream/blob/master/README.md"> <reference anchor="SCReAM-L4S" target="https://github.com/EricssonResear
ch/scream">
<front> <front>
<title>SCReAM</title> <title>SCReAM</title>
<author fullname="Ingemar Johansson" initials="I" surname="Johansson <author/>
"> <date month="June" year="2022"/>
<organization/>
</author>
<date/>
</front> </front>
<seriesInfo name="GitHub repository;" value=""/> <refcontent>commit fda6c53</refcontent>
<format target="https://github.com/google/bbr/blob/v2alpha/README.md"
type="Source code"/>
</reference> </reference>
<reference anchor="L4S_5G" target="https://www.ericsson.com/en/reports-a
nd-papers/white-papers/enabling-time-critical-applications-over-5g-with-rate-ada <reference anchor="L4S_5G" target="https://www.ericsson.com/en/reports-and-paper
ptation"> s/white-papers/enabling-time-critical-applications-over-5g-with-rate-adaptation"
>
<front> <front>
<title>Enabling time-critical applications over 5G with rate <title>Enabling time-critical applications over 5G with rate
adaptation</title> adaptation</title>
<author fullname="Per Willars" initials="P." surname="Willars"/> <author fullname="Per Willars" initials="P." surname="Willars"/>
<author fullname="Emma Wittenmark" initials="E." surname="Wittenmark "/> <author fullname="Emma Wittenmark" initials="E." surname="Wittenmark "/>
<author fullname="Henrik Ronkainen" initials="H." surname="Ronkainen "/> <author fullname="Henrik Ronkainen" initials="H." surname="Ronkainen "/>
<author fullname="Christer Östberg" initials="C." surname="Östberg"/ > <author fullname="Christer Östberg" initials="C." surname="Östberg"/ >
<author fullname="Ingemar Johansson" initials="I." surname="Johansso n"/> <author fullname="Ingemar Johansson" initials="I." surname="Johansso n"/>
<author fullname="Johan Strand" initials="J." surname="Strand"/> <author fullname="Johan Strand" initials="J." surname="Strand"/>
<author fullname="Petr Lédl" initials="P." surname="Lédl"/> <author fullname="Petr Lédl" initials="P." surname="Lédl"/>
<author fullname="Dominik Schnieders" initials="D." surname="Schnied ers"/> <author fullname="Dominik Schnieders" initials="D." surname="Schnied ers"/>
<date month="May" year="2021"/> <date month="May" year="2021"/>
</front> </front>
<seriesInfo name="Ericsson - Deutsche Telekom White Paper" value="BNEW -21:025455 Uen"/> <refcontent>Ericsson - Deutsche Telekom White Paper, BNEW-21:025455</r efcontent>
<format target="https://www.ericsson.com/49bc82/assets/local/reports-p apers/white-papers/26052021-enabling-time-critical-applications-over-5g-with-rat e-adaptation-whitepaper.pdf" type="PDF"/> <format target="https://www.ericsson.com/49bc82/assets/local/reports-p apers/white-papers/26052021-enabling-time-critical-applications-over-5g-with-rat e-adaptation-whitepaper.pdf" type="PDF"/>
</reference> </reference>
<reference anchor="BBRv2" target="https://github.com/google/bbr/blob/v2a
lpha/README.md"> <reference anchor="BBRv2" target="https://github.com/google/bbr">
<front> <front>
<title>BRTCP BBR v2 Alpha/Preview Release</title> <title>TCP BBR v2 Alpha/Preview Release</title>
<author fullname="Neal Cardwell" initials="N" surname="Cardwell"> <author/>
<organization/> <date month="June" year="2022"/>
</author>
<date/>
</front> </front>
<seriesInfo name="GitHub repository;" value="Linux congestion control module"/> <refcontent>commit 17700ca</refcontent>
</reference> </reference>
<reference anchor="Heist21" target="https://github.com/heistp/l4s-tests/
#underutilization-with-bursty-traffic"> <reference anchor="Heist21" target="https://github.com/heistp/l4s-tests"
>
<front> <front>
<title>L4S Tests</title> <title>L4S Tests</title>
<author fullname="Pete Heist" initials="P." surname="Heist"> <author/>
<organization/>
</author>
<author fullname="Jonathan Morton" initials="J." surname="Morton">
<organization/>
</author>
<date month="August" year="2021"/> <date month="August" year="2021"/>
</front> </front>
<seriesInfo name="GitHub" value="README"/> <refcontent>commit e21cd91</refcontent>
</reference> </reference>
<reference anchor="Boru20" target="https://dl.acm.org/doi/abs/10.1145/34 02413.3402419"> <reference anchor="Boru20" target="https://dl.acm.org/doi/abs/10.1145/34 02413.3402419">
<front> <front>
<title>Validating the Sharing Behavior and Latency Characteristics <title>Validating the Sharing Behavior and Latency Characteristics
of the L4S Architecture</title> of the L4S Architecture</title>
<author fullname="Dejene Boru Oljira" initials="D." surname="Boru Ol jira"> <author fullname="Dejene Boru Oljira" initials="D." surname="Boru Ol jira">
<organization>Karlstad Uni</organization> <organization>Karlstad Uni</organization>
</author> </author>
<author fullname="Karl-Johan Grinnemo" initials="K-J." surname="Grin nemo"> <author fullname="Karl-Johan Grinnemo" initials="K-J." surname="Grin nemo">
<organization>Karlstad Uni</organization> <organization>Karlstad Uni</organization>
</author> </author>
<author fullname="Anna Brunstrom" initials="A." surname="Brunstrom"> <author fullname="Anna Brunstrom" initials="A." surname="Brunstrom">
<organization>Karlstad Uni</organization> <organization>Karlstad Uni</organization>
</author> </author>
<author fullname="Javid Taheri" initials="J." surname="Taheri"> <author fullname="Javid Taheri" initials="J." surname="Taheri">
<organization>Karlstad Uni</organization> <organization>Karlstad Uni</organization>
</author> </author>
<date month="May" year="2020"/> <date month="May" year="2020"/>
</front> </front>
<seriesInfo name="ACM CCR" value="50(2):37--44"/> <seriesInfo name="DOI" value="10.1145/3402413.3402419"/>
<refcontent>ACM SIGCOMM Computer Communication Review, Vol. 50, Issue
2, pp. 37-44</refcontent>
</reference> </reference>
</references> </references>
</references> </references>
<section anchor="dualq_Ex_algo_pi2" numbered="true" toc="default"> <section anchor="dualq_Ex_algo_pi2" numbered="true" toc="default">
<name>Example DualQ Coupled PI2 Algorithm</name> <name>Example DualQ Coupled PI2 Algorithm</name>
<t>As a first concrete example, the pseudocode below gives the DualPI2 <t>As a first concrete example, the pseudocode below gives the DualPI2
algorithm. DualPI2 follows the structure of the DualQ Coupled AQM algorithm. DualPI2 follows the structure of the DualQ Coupled AQM
framework in <xref target="dualq_fig_structure" format="default"/>. A simp le ramp framework in <xref target="dualq_fig_structure" format="default"/>. A simp le ramp
function (configured in units of queuing time) with unsmoothed ECN function (configured in units of queuing time) with unsmoothed ECN
marking is used for the Native L4S AQM. The ramp can also be configured marking is used for the Native L4S AQM. The ramp can also be configured
as a step function. The PI2 algorithm <xref target="PI2" format="default"/ > is used as a step function. The PI2 algorithm <xref target="PI2" format="default"/ > is used
for the Classic AQM. PI2 is an improved variant of the PIE for the Classic AQM. PI2 is an improved variant of the PIE
AQM <xref target="RFC8033" format="default"/>.</t> AQM <xref target="RFC8033" format="default"/>.</t>
<t>The pseudocode will be introduced in two passes. The first pass <t>The pseudocode will be introduced in two passes. The first pass
explains the core concepts, deferring handling of edge-cases like explains the core concepts, deferring handling of edge-cases like
overload to the second pass. To aid comparison, line numbers are kept in overload to the second pass. To aid comparison, line numbers are kept in
step between the two passes by using letter suffixes where the longer step between the two passes by using letter suffixes where the longer
code needs extra lines.</t> code needs extra lines.</t>
<t>All variables are assumed to be floating point in their basic units <t>All variables are assumed to be floating point in their basic units
(size in bytes, time in seconds, rates in bytes/second, alpha and beta (size in bytes, time in seconds, rates in bytes/second, alpha and beta
in Hz, and probabilities from 0 to 1. Constants expressed in k (kilo), M in Hz, and probabilities from 0 to 1). Constants expressed in k (kilo), M
(mega), G (giga), u (micro), m (milli) , %, ... are assumed to be (mega), G (giga), u (micro), m (milli), %, and so forth, are assumed to be
converted to their appropriate multiple or fraction to represent the converted to their appropriate multiple or fraction to represent the
basic units. A real implementation that wants to use integer values basic units. A real implementation that wants to use integer values
needs to handle appropriate scaling factors and allow accordingly needs to handle appropriate scaling factors and allow
appropriate resolution of its integer types (including temporary appropriate resolution of its integer types (including temporary
internal values during calculations).</t> internal values during calculations).</t>
<t>A full open source implementation for Linux is available at: <t>A full open source implementation for Linux is available at
https://github.com/L4STeam/sch_dualpi2_upstream and explained in <xref tar <eref target="https://github.com/L4STeam/sch_dualpi2_upstream" brackets="a
get="DualPI2Linux" format="default"/>. The specification of the DualQ Coupled AQ ngle"/> and explained in <xref target="DualPI2Linux" format="default"/>. The spe
M for cification of the DualQ Coupled AQM for
DOCSIS cable modems and CMTSs is available in <xref target="DOCSIS3.1" for DOCSIS cable modems and cable modem termination systems (CMTSs) is availab
mat="default"/> le in <xref target="DOCSIS3.1" format="default"/>
and explained in <xref target="LLD" format="default"/>.</t> and explained in <xref target="LLD" format="default"/>.</t>
<section anchor="dualq_Ex_algo_pi2-1" numbered="true" toc="default"> <section anchor="dualq_Ex_algo_pi2-1" numbered="true" toc="default">
<name>Pass #1: Core Concepts</name> <name>Pass #1: Core Concepts</name>
<t>The pseudocode manipulates three main structures of variables: the <t>The pseudocode manipulates three main structures of variables: the
packet (pkt), the L4S queue (lq) and the Classic queue (cq). The packet (pkt), the L4S queue (lq), and the Classic queue (cq). The
pseudocode consists of the following six functions:</t> pseudocode consists of the following six functions:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>The initialization function dualpi2_params_init(...) (<xref target ="dualq_fig_Algo_pi2_core_header" format="default"/>) that sets parameter <li>The initialization function dualpi2_params_init(...) (<xref target ="dualq_fig_Algo_pi2_core_header" format="default"/>) that sets parameter
defaults (the API for setting non-default values is omitted for defaults (the API for setting non-default values is omitted for
brevity)</li> brevity).</li>
<li>The enqueue function dualpi2_enqueue(lq, cq, pkt) (<xref target="d <li>The enqueue function dualpi2_enqueue(lq, cq, pkt) (<xref target="d
ualq_fig_Algo_pi2_enqueue" format="default"/>)</li> ualq_fig_Algo_pi2_enqueue" format="default"/>).</li>
<li>The dequeue function dualpi2_dequeue(lq, cq, pkt) (<xref target="d <li>The dequeue function dualpi2_dequeue(lq, cq, pkt) (<xref target="d
ualq_fig_Algo_pi2_dequeue" format="default"/>)</li> ualq_fig_Algo_pi2_dequeue" format="default"/>).</li>
<li>The recurrence function recur(q, likelihood) for de-randomized <li>The recurrence function recur(q, likelihood) for de-randomized
ECN marking (shown at the end of <xref target="dualq_fig_Algo_pi2_de queue" format="default"/>).</li> ECN marking (shown at the end of <xref target="dualq_fig_Algo_pi2_de queue" format="default"/>).</li>
<li>The L4S AQM function laqm(qdelay) (<xref target="dualq_fig_Algo_la qm_core" format="default"/>) used to calculate the <li>The L4S AQM function laqm(qdelay) (<xref target="dualq_fig_Algo_la qm_core" format="default"/>) used to calculate the
ECN-marking probability for the L4S queue</li> ECN-marking probability for the L4S queue.</li>
<li>The base AQM function that implements the PI algorithm <li>The Base AQM function that implements the PI algorithm
dualpi2_update(lq, cq) (<xref target="dualq_fig_Algo_pi2_core" forma t="default"/>) dualpi2_update(lq, cq) (<xref target="dualq_fig_Algo_pi2_core" forma t="default"/>)
used to regularly update the base probability (p'), which is used to regularly update the base probability (p'), which is
squared for the Classic AQM as well as being coupled across to the squared for the Classic AQM as well as being coupled across to the
L4S queue.</li> L4S queue.</li>
</ul> </ul>
<t>It also uses the following functions that are not shown in <t>It also uses the following functions that are not shown in
full here:</t> full here:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>scheduler(), which selects between the head packets of the two <li>scheduler(), which selects between the head packets of the two
queues; the choice of scheduler technology is discussed later;</li> queues. The choice of scheduler technology is discussed later.</li>
<li>cq.byt() or lq.byt() returns the current length <li>cq.byt() or lq.byt() returns the current length
(aka. backlog) of the relevant queue in bytes;</li> (a.k.a. backlog) of the relevant queue in bytes.</li>
<li>cq.len() or lq.len() returns the current length of the relevant <li>cq.len() or lq.len() returns the current length of the relevant
queue in packets;</li> queue in packets.</li>
<li>cq.time() or lq.time() returns the current queuing delay of the <li>cq.time() or lq.time() returns the current queuing delay of the
relevant queue in units of time (see Note a);</li> relevant queue in units of time (see <xref target="note_qdelay" form
<li>mark(pkt) and drop(pkt) for ECN-marking and dropping a at="none">Note a</xref> below).</li>
packet;</li> <li>mark(pkt) and drop(pkt) for ECN marking and dropping a
packet.</li>
</ul> </ul>
<t>In experiments so far (building on experiments with PIE) on <t>In experiments so far (building on experiments with PIE) on
broadband access links ranging from 4 Mb/s to 200 Mb/s with base RTTs broadband access links ranging from 4 Mb/s to 200 Mb/s with base RTTs
from 5 ms to 100 ms, DualPI2 achieves good results with the default from 5 ms to 100 ms, DualPI2 achieves good results with the default
parameters in <xref target="dualq_fig_Algo_pi2_core_header" format="defa ult"/>. The parameters in <xref target="dualq_fig_Algo_pi2_core_header" format="defa ult"/>. The
parameters are categorised by whether they relate to the Base PI2 AQM, parameters are categorised by whether they relate to the PI2 AQM,
the L4S AQM or the framework coupling them together. Constants and the L4S AQM, or the framework coupling them together. Constants and
variables derived from these parameters are also included at the end variables derived from these parameters are also included at the end
of each category. Each parameter is explained as it is encountered in of each category. Each parameter is explained as it is encountered in
the walk-through of the pseudocode below, and the rationale for the the walk-through of the pseudocode below, and the rationale for the
chosen defaults are given so that sensible values can be used in chosen defaults are given so that sensible values can be used in
scenarios other than the regular public Internet.</t> scenarios other than the regular public Internet.</t>
<figure anchor="dualq_fig_Algo_pi2_core_header"> <figure anchor="dualq_fig_Algo_pi2_core_header">
<name>Example Header Pseudocode for DualQ Coupled PI2 AQM</name> <name>Example Header Pseudocode for DualQ Coupled PI2 AQM</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: dualpi2_para <sourcecode><![CDATA[
ms_init(...) { % Set input parameter defaults 1: dualpi2_params_init(...) { % Set input parameter defaults
2: % DualQ Coupled framework parameters 2: % DualQ Coupled framework parameters
5: limit = MAX_LINK_RATE * 250 ms % Dual buffer size 5: limit = MAX_LINK_RATE * 250 ms % Dual buffer size
3: k = 2 % Coupling factor 3: k = 2 % Coupling factor
4: % NOT SHOWN % scheduler-dependent weight or equival't parameter 4: % NOT SHOWN % scheduler-dependent weight or equival't parameter
6: 6:
7: % PI2 Classic AQM parameters 7: % PI2 Classic AQM parameters
8: target = 15 ms % Queue delay target 8: target = 15 ms % Queue delay target
9: RTT_max = 100 ms % Worst case RTT expected 9: RTT_max = 100 ms % Worst case RTT expected
10: % PI2 constants derived from above PI2 parameters 10: % PI2 constants derived from above PI2 parameters
11: p_Cmax = min(1/k^2, 1) % Max Classic drop/mark prob 11: p_Cmax = min(1/k^2, 1) % Max Classic drop/mark prob
12: Tupdate = min(target, RTT_max/3) % PI sampling interval 12: Tupdate = min(target, RTT_max/3) % PI sampling interval
13: alpha = 0.1 * Tupdate / RTT_max^2 % PI integral gain in Hz 13: alpha = 0.1 * Tupdate / RTT_max^2 % PI integral gain in Hz
14: beta = 0.3 / RTT_max % PI proportional gain in Hz 14: beta = 0.3 / RTT_max % PI proportional gain in Hz
15: 15:
16: % L4S ramp AQM parameters 16: % L4S ramp AQM parameters
17: minTh = 800 us % L4S min marking threshold in time units 17: minTh = 800 us % L4S min marking threshold in time units
18: range = 400 us % Range of L4S ramp in time units 18: range = 400 us % Range of L4S ramp in time units
19: Th_len = 1 pkt % Min L4S marking threshold in packets 19: Th_len = 1 pkt % Min L4S marking threshold in packets
20: % L4S constants 20: % L4S constants
21: p_Lmax = 1 % Max L4S marking prob 21: p_Lmax = 1 % Max L4S marking prob
22: } 22: }]]></sourcecode>
]]></artwork>
</figure> </figure>
<t>The overall goal of the code is to apply the marking and dropping <t>The overall goal of the code is to apply the marking and dropping
probabilities for L4S and Classic traffic (p_L and p_C). These are probabilities for L4S and Classic traffic (p_L and p_C). These are
derived from the underlying base probabilities p'_L and p' driven derived from the underlying base probabilities p'_L and p' driven,
respectively by the traffic in the L and C queues. The marking respectively, by the traffic in the L and C queues. The marking
probability for the L queue (p_L) depends on both the base probability probability for the L queue (p_L) depends on both the base probability
in its own queue (p'_L) and a probability called p_CL, which is in its own queue (p'_L) and a probability called p_CL, which is
coupled across from p' in the C queue (see <xref target="dualq_coupled_s tructure" format="default"/> for the derivation of the specific coupled across from p' in the C queue (see <xref target="dualq_coupled_s tructure" format="default"/> for the derivation of the specific
equations and dependencies).</t> equations and dependencies).</t>
<t>The probabilities p_CL and p_C are derived in lines 4 and 5 of the <t>The probabilities p_CL and p_C are derived in lines 4 and 5 of the
dualpi2_update() function (<xref target="dualq_fig_Algo_pi2_core" format ="default"/>) dualpi2_update() function (<xref target="dualq_fig_Algo_pi2_core" format ="default"/>)
then used in the dualpi2_dequeue() function where p_L is also derived then used in the dualpi2_dequeue() function where p_L is also derived
from p_CL at line 6 (<xref target="dualq_fig_Algo_pi2_dequeue" format="d efault"/>). The from p_CL at line 6 (<xref target="dualq_fig_Algo_pi2_dequeue" format="d efault"/>). The
code walk-through below builds up to explaining that part of the code code walk-through below builds up to explaining that part of the code
eventually, but it starts from packet arrival.</t> eventually, but it starts from packet arrival.</t>
<figure anchor="dualq_fig_Algo_pi2_enqueue"> <figure anchor="dualq_fig_Algo_pi2_enqueue">
<name>Example Enqueue Pseudocode for DualQ Coupled PI2 AQM</name> <name>Example Enqueue Pseudocode for DualQ Coupled PI2 AQM</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: dualpi2_enqu <sourcecode><![CDATA[
eue(lq, cq, pkt) { % Test limit and classify lq or cq 1: dualpi2_enqueue(lq, cq, pkt) { % Test limit and classify lq or cq
2: if ( lq.byt() + cq.byt() + MTU > limit) 2: if ( lq.byt() + cq.byt() + MTU > limit)
3: drop(pkt) % drop packet if buffer is full 3: drop(pkt) % drop packet if buffer is full
4: timestamp(pkt) % only needed if using the sojourn technique 4: timestamp(pkt) % only needed if using the sojourn technique
5: % Packet classifier 5: % Packet classifier
6: if ( ecn(pkt) modulo 2 == 1 ) % ECN bits = ECT(1) or CE 6: if ( ecn(pkt) modulo 2 == 1 ) % ECN bits = ECT(1) or CE
7: lq.enqueue(pkt) 7: lq.enqueue(pkt)
8: else % ECN bits = not-ECT or ECT(0) 8: else % ECN bits = not-ECT or ECT(0)
9: cq.enqueue(pkt) 9: cq.enqueue(pkt)
10: } 10: }]]></sourcecode>
]]></artwork>
</figure> </figure>
<figure anchor="dualq_fig_Algo_pi2_dequeue"> <figure anchor="dualq_fig_Algo_pi2_dequeue">
<name>Example Dequeue Pseudocode for DualQ Coupled PI2 AQM</name> <name>Example Dequeue Pseudocode for DualQ Coupled PI2 AQM</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: dualpi2_dequ <sourcecode><![CDATA[
eue(lq, cq, pkt) { % Couples L4S & Classic queues 1: dualpi2_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: while ( lq.byt() + cq.byt() > 0 ) { 2: while ( lq.byt() + cq.byt() > 0 ) {
3: if ( scheduler() == lq ) { 3: if ( scheduler() == lq ) {
4: lq.dequeue(pkt) % Scheduler chooses lq 4: lq.dequeue(pkt) % Scheduler chooses lq
5: p'_L = laqm(lq.time()) % Native LAQM 5: p'_L = laqm(lq.time()) % Native LAQM
6: p_L = max(p'_L, p_CL) % Combining function 6: p_L = max(p'_L, p_CL) % Combining function
7: if ( recur(lq, p_L) ) % Linear marking 7: if ( recur(lq, p_L) ) % Linear marking
8: mark(pkt) 8: mark(pkt)
9: } else { 9: } else {
10: cq.dequeue(pkt) % Scheduler chooses cq 10: cq.dequeue(pkt) % Scheduler chooses cq
11: if ( recur(cq, p_C) ) { % probability p_C = p'^2 11: if ( recur(cq, p_C) ) { % probability p_C = p'^2
skipping to change at line 2607 skipping to change at line 1982
21: return(NULL) % no packet to dequeue 21: return(NULL) % no packet to dequeue
22: } 22: }
23: recur(q, likelihood) { % Returns TRUE with a certain likelihood 23: recur(q, likelihood) { % Returns TRUE with a certain likelihood
24: q.count += likelihood 24: q.count += likelihood
25: if (q.count > 1) { 25: if (q.count > 1) {
26: q.count -= 1 26: q.count -= 1
27: return TRUE 27: return TRUE
28: } 28: }
29: return FALSE 29: return FALSE
30: } 30: }]]></sourcecode>
]]></artwork>
</figure> </figure>
<t>When packets arrive, first a common queue limit is checked as shown <t>When packets arrive, a common queue limit is checked first as shown
in line 2 of the enqueuing pseudocode in <xref target="dualq_fig_Algo_pi 2_enqueue" format="default"/>. This assumes a shared buffer in line 2 of the enqueuing pseudocode in <xref target="dualq_fig_Algo_pi 2_enqueue" format="default"/>. This assumes a shared buffer
for the two queues (Note b discusses the merits of separate buffers). for the two queues (<xref target="note_separate_buffers" format="none">N ote b</xref> discusses the merits of separate buffers).
In order to avoid any bias against larger packets, 1 MTU of space is In order to avoid any bias against larger packets, 1 MTU of space is
always allowed, and the limit is deliberately tested before always allowed, and the limit is deliberately tested before
enqueue.</t> enqueue.</t>
<t>If limit is not exceeded, the packet is timestamped in line 4 (only <t>If limit is not exceeded, the packet is timestamped in line 4 (only
if the sojourn time technique is being used to measure queue delay; if the sojourn time technique is being used to measure queue delay;
see Note a for alternatives).</t> see <xref target="note_qdelay" format="none">Note a</xref> below for alt ernatives).</t>
<t>At lines 5-9, the packet is classified and enqueued to the Classic <t>At lines 5-9, the packet is classified and enqueued to the Classic
or L4S queue dependent on the least significant bit of the ECN field or L4S queue dependent on the least significant bit (LSB) of the ECN fie ld
in the IP header (line 6). Packets with a codepoint having an LSB of 0 in the IP header (line 6). Packets with a codepoint having an LSB of 0
(Not-ECT and ECT(0)) will be enqueued in the Classic queue. Otherwise, (Not-ECT and ECT(0)) will be enqueued in the Classic queue. Otherwise,
ECT(1) and CE packets will be enqueued in the L4S queue. Optional ECT(1) and CE packets will be enqueued in the L4S queue. Optional
additional packet classification flexibility is omitted for brevity additional packet classification flexibility is omitted for brevity
(see the L4S ECN protocol <xref target="I-D.ietf-tsvwg-ecn-l4s-id" forma t="default"/>).</t> (see the L4S ECN protocol <xref target="RFC9331" format="default"/>).</t >
<t>The dequeue pseudocode (<xref target="dualq_fig_Algo_pi2_dequeue" for mat="default"/>) is repeatedly called whenever <t>The dequeue pseudocode (<xref target="dualq_fig_Algo_pi2_dequeue" for mat="default"/>) is repeatedly called whenever
the lower layer is ready to forward a packet. It schedules one packet the lower layer is ready to forward a packet. It schedules one packet
for dequeuing (or zero if the queue is empty) then returns control to for dequeuing (or zero if the queue is empty) then returns control to
the caller, so that it does not block while that packet is being the caller so that it does not block while that packet is being
forwarded. While making this dequeue decision, it also makes the forwarded. While making this dequeue decision, it also makes the
necessary AQM decisions on dropping or marking. The alternative of necessary AQM decisions on dropping or marking. The alternative of
applying the AQMs at enqueue would shift some processing from the applying the AQMs at enqueue would shift some processing from the
critical time when each packet is dequeued. However, it would also add critical time when each packet is dequeued. However, it would also add
a whole queue of delay to the control signals, making the control loop a whole queue of delay to the control signals, making the control loop
sloppier (for a typical RTT it would double the Classic queue's sloppier (for a typical RTT, it would double the Classic queue's
feedback delay).</t> feedback delay).</t>
<t>All the dequeue code is contained within a large while loop so that <t>All the dequeue code is contained within a large while loop so that
if it decides to drop a packet, it will continue until it selects a if it decides to drop a packet, it will continue until it selects a
packet to schedule. Line 3 of the dequeue pseudocode is where the packet to schedule. Line 3 of the dequeue pseudocode is where the
scheduler chooses between the L4S queue (lq) and the Classic queue scheduler chooses between the L4S queue (lq) and the Classic queue
(cq). Detailed implementation of the scheduler is not shown (see (cq). Detailed implementation of the scheduler is not shown (see
discussion later). </t> discussion later). </t>
<ul spacing="normal"> <ul spacing="normal">
<li> <li>
<t>If an L4S packet is scheduled, in lines 7 and 8 the packet is <t>If an L4S packet is scheduled, in lines 7 and 8 the packet is
ECN-marked with likelihood p_L. The recur() function at the end of ECN-marked with likelihood p_L. The recur() function at the end of
<xref target="dualq_fig_Algo_pi2_dequeue" format="default"/> is used , which is <xref target="dualq_fig_Algo_pi2_dequeue" format="default"/> is used , which is
preferred over random marking because it avoids delay due to preferred over random marking because it avoids delay due to
randomization when interpreting congestion signals, but it still randomization when interpreting congestion signals, but it still
desynchronizes the saw-teeth of the flows. Line 6 calculates p_L desynchronizes the sawteeth of the flows. Line 6 calculates p_L
as the maximum of the coupled L4S probability p_CL and the as the maximum of the coupled L4S probability p_CL and the
probability from the native L4S AQM p'_L. This implements the probability from the Native L4S AQM p'_L. This implements the
max() function shown in <xref target="dualq_fig_structure" format="d efault"/> to max() function shown in <xref target="dualq_fig_structure" format="d efault"/> to
couple the outputs of the two AQMs together. Of the two couple the outputs of the two AQMs together. Of the two
probabilities input to p_L in line 6:</t> probabilities input to p_L in line 6:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>p'_L is calculated per packet in line 5 by the laqm() <li>p'_L is calculated per packet in line 5 by the laqm()
function (see <xref target="dualq_fig_Algo_laqm_core" format="de function (see <xref target="dualq_fig_Algo_laqm_core" format="de
fault"/>),</li> fault"/>), whereas</li>
<li>Whereas p_CL is maintained by the dualpi2_update() function <li>p_CL is maintained by the dualpi2_update() function,
which runs every Tupdate (Tupdate is set in line 12 of <xref tar get="dualq_fig_Algo_pi2_core_header" format="default"/>).</li> which runs every Tupdate (Tupdate is set in line 12 of <xref tar get="dualq_fig_Algo_pi2_core_header" format="default"/>).</li>
</ul> </ul>
</li> </li>
<li>If a Classic packet is scheduled, lines 10 to 17 drop or mark <li>If a Classic packet is scheduled, lines 10 to 17 drop or mark
the packet with probability p_C.</li> the packet with probability p_C.</li>
</ul> </ul>
<t>The Native L4S AQM algorithm (<xref target="dualq_fig_Algo_laqm_core" format="default"/>) is a ramp function, similar to <t>The Native L4S AQM algorithm (<xref target="dualq_fig_Algo_laqm_core" format="default"/>) is a ramp function, similar to
the RED algorithm, but simplified as follows:</t> the RED algorithm, but simplified as follows:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>The extent of the ramp is defined in units of queuing delay, <li>The extent of the ramp is defined in units of queuing delay,
not bytes, so that configuration remains invariant as the queue not bytes, so that configuration remains invariant as the queue
departure rate varies.</li> departure rate varies.</li>
<li>It uses instantaneous queueing delay, which avoids the <li>It uses instantaneous queuing delay, which avoids the
complexity of smoothing, but also avoids embedding a worst-case complexity of smoothing, but also avoids embedding a worst-case
RTT of smoothing delay in the network (see <xref target="dualq_coupl ed" format="default"/>).</li> RTT of smoothing delay in the network (see <xref target="dualq_coupl ed" format="default"/>).</li>
<li>The ramp rises linearly directly from 0 to 1, not to an <li>The ramp rises linearly directly from 0 to 1, not to an
intermediate value of p'_L as RED would, because there is no need intermediate value of p'_L as RED would, because there is no need
to keep ECN marking probability low.</li> to keep ECN-marking probability low.</li>
<li>Marking does not have to be randomized. Determinism is used <li>Marking does not have to be randomized. Determinism is used
instead of randomness; to reduce the delay necessary to smooth out instead of randomness to reduce the delay necessary to smooth out
the noise of randomness from the signal.</li> the noise of randomness from the signal.</li>
</ul> </ul>
<t>The ramp function requires two configuration parameters, the <t>The ramp function requires two configuration parameters, the
minimum threshold (minTh) and the width of the ramp (range), both in minimum threshold (minTh) and the width of the ramp (range), both in
units of queuing time, as shown in lines 17 &amp; 18 of the units of queuing time, as shown in lines 17 and 18 of the
initialization function in <xref target="dualq_fig_Algo_pi2_core_header" format="default"/>. The ramp function can be initialization function in <xref target="dualq_fig_Algo_pi2_core_header" format="default"/>. The ramp function can be
configured as a step (see Note c).</t> configured as a step (see <xref target="note_ramp" format="none">Note c<
<t>Although the DCTCP paper <xref target="Alizadeh-stability" format="de /xref>).</t>
fault"/> <t>Although the DCTCP paper <xref target="Alizadeh-stability" format="de
recommends an ECN marking threshold of 0.17*RTT_typ, it also shows fault"/>
recommends an ECN-marking threshold of 0.17*RTT_typ, it also shows
that the threshold can be much shallower with hardly any worse that the threshold can be much shallower with hardly any worse
under-utilization of the link (because the amplitude of DCTCP's underutilization of the link (because the amplitude of DCTCP's
sawteeth is so small). Based on extensive experiments, for the public sawteeth is so small). Based on extensive experiments, for the public
Internet the default minimum ECN marking threshold (target) in <xref tar Internet the default minimum ECN-marking threshold (target) in <xref tar
get="dualq_fig_Algo_pi2_core_header" format="default"/> is considered a good get="dualq_fig_Algo_pi2_core_header" format="default"/> is considered a good
compromise, even though it is significantly smaller fraction of compromise, even though it is a significantly smaller fraction of
RTT_typ.</t> RTT_typ.</t>
<figure anchor="dualq_fig_Algo_laqm_core"> <figure anchor="dualq_fig_Algo_laqm_core">
<name>Example Pseudocode for the Native L4S AQM</name> <name>Example Pseudocode for the Native L4S AQM</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: laqm(qdelay) <sourcecode><![CDATA[
{ % Returns native L4S AQM probability 1: laqm(qdelay) { % Returns Native L4S AQM probability
2: if (qdelay >= maxTh) 2: if (qdelay >= maxTh)
3: return 1 3: return 1
4: else if (qdelay > minTh) 4: else if (qdelay > minTh)
5: return (qdelay - minTh)/range % Divide could use a bit-shift 5: return (qdelay - minTh)/range % Divide could use a bit-shift
6: else 6: else
7: return 0 7: return 0
8: } 8: }]]></sourcecode>
]]></artwork>
</figure> </figure>
<t/> <t/>
<figure anchor="dualq_fig_Algo_pi2_core"> <figure anchor="dualq_fig_Algo_pi2_core">
<name>Example PI-Update Pseudocode for DualQ Coupled PI2 AQM</name> <name>Example PI-update Pseudocode for DualQ Coupled PI2 AQM</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: dualpi2_upda <sourcecode><![CDATA[
te(lq, cq) { % Update p' every Tupdate 1: dualpi2_update(lq, cq) { % Update p' every Tupdate
2: curq = cq.time() % use queuing time of first-in Classic packet 2: curq = cq.time() % use queuing time of first-in Classic packet
3: p' = p' + alpha * (curq - target) + beta * (curq - prevq) 3: p' = p' + alpha * (curq - target) + beta * (curq - prevq)
4: p_CL = k * p' % Coupled L4S prob = base prob * coupling factor 4: p_CL = k * p' % Coupled L4S prob = base prob * coupling factor
5: p_C = p'^2 % Classic prob = (base prob)^2 5: p_C = p'^2 % Classic prob = (base prob)^2
6: prevq = curq 6: prevq = curq
7: } 7: }]]></sourcecode>
]]></artwork>
</figure> </figure>
<t keepWithPrevious="true">(Clamping p' within the range [0,1] omitted f <t keepWithPrevious="true" indent='3'>(Note: Clamping p' within the rang
or clarity - e [0,1] omitted for clarity -- see below.)</t>
see text)</t> <t>The coupled marking probability p_CL depends on the base
<t>The coupled marking probability, p_CL depends on the base probability (p'), which is kept up to date by executing the core PI algo
probability (p'), which is kept up to date by the core PI algorithm in rithm in
<xref target="dualq_fig_Algo_pi2_core" format="default"/> executed every <xref target="dualq_fig_Algo_pi2_core" format="default"/> every Tupdate.
Tupdate.</t> </t>
<t>Note that p' solely depends on the queuing time in the Classic <t>Note that p' solely depends on the queuing time in the Classic
queue. In line 2, the current queuing delay (curq) is evaluated from queue. In line 2, the current queuing delay (curq) is evaluated from
how long the head packet was in the Classic queue (cq). The function how long the head packet was in the Classic queue (cq). The function
cq.time() (not shown) subtracts the time stamped at enqueue from the cq.time() (not shown) subtracts the time stamped at enqueue from the
current time (see Note a) and implicitly takes the current queuing current time (see <xref target="note_qdelay"
format="none">Note a</xref> below) and implicitly takes the current queuing
delay as 0 if the queue is empty.</t> delay as 0 if the queue is empty.</t>
<t>The algorithm centres on line 3, which is a classical <t>The algorithm centres on line 3, which is a classical
Proportional-Integral (PI) controller that alters p' dependent on: a) PI controller that alters p' dependent on: a)
the error between the current queuing delay (curq) and the target the error between the current queuing delay (curq) and the target
queuing delay, 'target'; and b) the change in queuing delay since the queuing delay (target) and b) the change in queuing delay since the
last sample. The name 'PI' represents the fact that the second factor last sample. The name 'PI' represents the fact that the second factor
(how fast the queue is growing) is <em>P</em>roportional (how fast the queue is growing) is Proportional
to load while the first is the <em>I</em>ntegral of to load while the first is the Integral of
the load (so it removes any standing queue in excess of the the load (so it removes any standing queue in excess of the
target).</t> target).</t>
<t>The target parameter can be set based on local knowledge, but the <t>The target parameter can be set based on local knowledge, but the
aim is for the default to be a good compromise for anywhere in the aim is for the default to be a good compromise for anywhere in the
intended deployment environment -- the public Internet. According intended deployment environment -- the public Internet. According
to <xref target="PI2param" format="default"/>, the target queuing delay on line 9 of to <xref target="PI2param" format="default"/>, the target queuing delay on line 8 of
<xref target="dualq_fig_Algo_pi2_core_header" format="default"/> is rela ted to the <xref target="dualq_fig_Algo_pi2_core_header" format="default"/> is rela ted to the
typical base RTT worldwide, RTT_typ, by two factors: target = RTT_typ typical base RTT worldwide, RTT_typ, by two factors: target = RTT_typ
* g * f. Below we summarize the rationale behind these factors and * g * f. Below, we summarize the rationale behind these factors and
introduce a further adjustment. The two factors ensure that, in a introduce a further adjustment. The two factors ensure that, in a
large proportion of cases (say 90%), the sawtooth variations in RTT of large proportion of cases (say 90%), the sawtooth variations in RTT of
a single flow will fit within the buffer without underutilizing the a single flow will fit within the buffer without underutilizing the
link. Frankly, these factors are educated guesses, but with the link. Frankly, these factors are educated guesses, but with the
emphasis closer to 'educated' than to 'guess' (see <xref target="PI2para m" format="default"/> for full background):</t> emphasis closer to 'educated' than to 'guess' (see <xref target="PI2para m" format="default"/> for the full background):</t>
<ul spacing="normal"> <ul spacing="normal">
<li>RTT_typ is taken as 25 ms. This is based on an average CDN <li>RTT_typ is taken as 25 ms. This is based on an average CDN
latency measured in each country weighted by the number of latency measured in each country weighted by the number of
Internet users in that country to produce an overall weighted Internet users in that country to produce an overall weighted
average for the Internet <xref target="PI2param" format="default"/>. Countries average for the Internet <xref target="PI2param" format="default"/>. Countries
were ranked by number of Internet users, and once 90% of Internet were ranked by number of Internet users, and once 90% of Internet
users were covered, smaller countries were excluded to avoid users were covered, smaller countries were excluded to avoid
unrepresentatively small sample sizes. Also, importantly, the data small sample sizes that would be less representative. Also, importan tly, the data
for the average CDN latency in China (with the largest number of for the average CDN latency in China (with the largest number of
Internet users) has been removed, because the CDN latency was a Internet users) has been removed, because the CDN latency was a
significant outlier and, on reflection, the experimental technique significant outlier and, on reflection, the experimental technique
seemed inappropriate to the CDN market in China.</li> seemed inappropriate to the CDN market in China.</li>
<li>g is taken as 0.38. The factor g is a geometry factor that <li>g is taken as 0.38. The factor g is a geometry factor that
characterizes the shape of the sawteeth of prevalent Classic characterizes the shape of the sawteeth of prevalent Classic
congestion controllers. The geometry factor is the fraction of the congestion controllers. The geometry factor is the fraction of the
amplitude of the sawtooth variability in queue delay that lies amplitude of the sawtooth variability in queue delay that lies
below the AQM's target. For instance, at low bit rate, the below the AQM's target.
geometry factor of standard Reno is 0.5, but at higher rates it For instance, at low bitrates, the
tends to just under 1. According to the census of congestion geometry factor of standard Reno is 0.5, but at higher rates, it
controllers conducted by Mishra et al. in Jul-Oct tends towards just under 1. According to the census of congestion
2019 <xref target="CCcensus19" format="default"/>, most Classic TCP controllers conducted by Mishra et al. in Jul-Oct
traffic 2019 <xref target="CCcensus19" format="default"/>, most Classic TCP
uses Cubic. And, according to the analysis in <xref target="PI2param traffic
" format="default"/>, if running over a PI2 AQM, a large proportion uses CUBIC. And, according to the analysis in <xref target="PI2param
of this Cubic traffic would be in its Reno-Friendly mode, which " format="default"/>, if running over a PI2 AQM, a large proportion
has a geometry factor of ~0.39 (all known implementations). The of this CUBIC traffic would be in its Reno-friendly mode, which
rest of the Cubic traffic would be in true Cubic mode, which has a has a geometry factor of ~0.39 (for all known implementations). The
rest of the CUBIC traffic would be in true CUBIC mode, which has a
geometry factor of ~0.36. Without modelling the sawtooth profiles geometry factor of ~0.36. Without modelling the sawtooth profiles
from all the other less prevalent congestion controllers, we from all the other less prevalent congestion controllers, we
estimate a 7:3 weighted average of these two, resulting in an estimate a 7:3 weighted average of these two, resulting in an
average geometry factor of 0.38.</li> average geometry factor of 0.38.</li>
<li>f is taken as 2. The factor f is a safety factor that increases <li>f is taken as 2. The factor f is a safety factor that increases
the target queue to allow for the distribution of RTT_typ around the target queue to allow for the distribution of RTT_typ around
its mean. Otherwise, the target queue would only avoid its mean. Otherwise, the target queue would only avoid
underutilization for those users below the mean. It also provides underutilization for those users below the mean. It also provides
a safety margin for the proportion of paths in use that span a safety margin for the proportion of paths in use that span
beyond the distance between a user and their local CDN. Currently, beyond the distance between a user and their local CDN. Currently,
no data is available on the variance of queue delay around the no data is available on the variance of queue delay around the
mean in each region, so there is plenty of room for this guess to mean in each region, so there is plenty of room for this guess to
become more educated.</li> become more educated.</li>
<li> <li>
<xref target="PI2param" format="default"/> recommends target = RTT_t yp * g * f = <xref target="PI2param" format="default"/> recommends target = RTT_t yp * g * f =
25ms * 0.38 * 2 = 19 ms. However, a further adjustment is 25 ms * 0.38 * 2 = 19 ms. However, a further adjustment is
warranted, because target is moving year-on-year. The paper is warranted, because target is moving year-on-year.
based on data collected in 2019, and it mentions evidence from The paper is
speedtest.net that suggests RTT_typ reduced by 17% (fixed) or 12% based on data collected in 2019, and it mentions evidence from the S
peedtest Global Index
that suggests RTT_typ reduced by 17% (fixed) or 12%
(mobile) between 2020 and 2021. Therefore, we recommend a default (mobile) between 2020 and 2021. Therefore, we recommend a default
of target = 15 ms at the time of writing (2021).</li> of target = 15 ms at the time of writing (2021).</li>
</ul> </ul>
<t>Operators can always use the data and discussion in <xref target="PI2 param" format="default"/> to configure a more appropriate target for their <t>Operators can always use the data and discussion in <xref target="PI2 param" format="default"/> to configure a more appropriate target for their
environment. For instance, an operator might wish to question the environment. For instance, an operator might wish to question the
assumptions called out in that paper, such as the goal of no assumptions called out in that paper, such as the goal of no
underutilization for a large majority of single flow transfers (given underutilization for a large majority of single flow transfers (given
many large transfers use multiple flows to avoid the scaling many large transfers use multiple flows to avoid the scaling
limitations of Classic flows).</t> limitations of Classic flows).</t>
<t>The two 'gain factors' in line 3 of <xref target="dualq_fig_Algo_pi2_ core" format="default"/>, alpha and beta, respectively <t>The two 'gain factors' in line 3 of <xref target="dualq_fig_Algo_pi2_ core" format="default"/>, alpha and beta, respectively
weight how strongly each of the two elements (Integral and weight how strongly each of the two elements (Integral and
Proportional) alters p'. They are in units of 'per second of delay' or Proportional) alters p'. They are in units of 'per second of delay' or
Hz, because they transform differences in queueing delay into changes Hz, because they transform differences in queuing delay into changes
in probability (assuming probability has a value from 0 to 1).</t> in probability (assuming probability has a value from 0 to 1).</t>
<t>Alpha and beta determine how much p' ought to change after each <t>Alpha and beta determine how much p' ought to change after each
update interval (Tupdate). For smaller Tupdate, p' should change by update interval (Tupdate). For a smaller Tupdate, p' should change by
the same amount per second, but in finer more frequent steps. So alpha the same amount per second but in finer more frequent steps. So alpha
depends on Tupdate (see line 13 of the initialization function in depends on Tupdate (see line 13 of the initialization function in
<xref target="dualq_fig_Algo_pi2_core_header" format="default"/>). It is best to update <xref target="dualq_fig_Algo_pi2_core_header" format="default"/>). It is best to update
p' as frequently as possible, but Tupdate will probably be constrained p' as frequently as possible, but Tupdate will probably be constrained
by hardware performance. As shown in line 13, the update interval by hardware performance. As shown in line 12, the update interval
should be frequent enough to update at least once in the time taken should be frequent enough to update at least once in the time taken
for the target queue to drain ('target') as long as it updates at for the target queue to drain ('target') as long as it updates at
least three times per maximum RTT. Tupdate defaults to 16 ms in the least three times per maximum RTT. Tupdate defaults to 16 ms in the
reference Linux implementation because it has to be rounded to a reference Linux implementation because it has to be rounded to a
multiple of 4 ms. For link rates from 4 to 200 Mb/s and a maximum RTT multiple of 4 ms. For link rates from 4 to 200 Mb/s and a maximum RTT
of 100ms, it has been verified through extensive testing that of 100 ms, it has been verified through extensive testing that
Tupdate=16ms (as also recommended in the PIE spec <xref target="RFC8033" Tupdate = 16 ms (as also recommended in the PIE spec <xref target="RFC80
format="default"/>) is sufficient.</t> 33" format="default"/>) is sufficient.</t>
<t>The choice of alpha and beta also determines the AQM's stable <t>The choice of alpha and beta also determines the AQM's stable
operating range. The AQM ought to change p' as fast as possible in operating range. The AQM ought to change p' as fast as possible in
response to changes in load without over-compensating and therefore response to changes in load without overcompensating and therefore
causing oscillations in the queue. Therefore, the values of alpha and causing oscillations in the queue. Therefore, the values of alpha and
beta also depend on the RTT of the expected worst-case flow beta also depend on the RTT of the expected worst-case flow
(RTT_max).</t> (RTT_max).</t>
<t>The maximum RTT of a PI controller (RTT_max in line 10 of <xref targe t="dualq_fig_Algo_pi2_core_header" format="default"/>) is not an absolute maximu m, <t>The maximum RTT of a PI controller (RTT_max in line 9 of <xref target ="dualq_fig_Algo_pi2_core_header" format="default"/>) is not an absolute maximum ,
but more instability (more queue variability) sets in for long-running but more instability (more queue variability) sets in for long-running
flows with an RTT above this value. The propagation delay halfway flows with an RTT above this value. The propagation delay halfway
round the planet and back in glass fibre is 200 ms. However, hardly round the planet and back in glass fibre is 200 ms. However, hardly
any traffic traverses such extreme paths and, since the significant any traffic traverses such extreme paths and, since the significant
consolidation of Internet traffic between 2007 and 2009 <xref target="La bovitz10" format="default"/>, a high and growing proportion of all Internet consolidation of Internet traffic between 2007 and 2009 <xref target="La bovitz10" format="default"/>, a high and growing proportion of all Internet
traffic (roughly two-thirds at the time of writing) has been served traffic (roughly two-thirds at the time of writing) has been served
from content distribution networks (CDNs) or 'cloud' services from CDNs or 'cloud' services
distributed close to end-users. The Internet might change again, but distributed close to end users. The Internet might change again, but
for now, designing for a maximum RTT of 100ms is a good compromise for now, designing for a maximum RTT of 100 ms is a good compromise
between faster queue control at low RTT and some instability on the between faster queue control at low RTT and some instability on the
occasions when a longer path is necessary.</t> occasions when a longer path is necessary.</t>
<t>Recommended derivations of the gain constants alpha and beta can be <t>Recommended derivations of the gain constants alpha and beta can be
approximated for Reno over a PI2 AQM as: alpha = 0.1 * Tupdate / approximated for Reno over a PI2 AQM as:
RTT_max^2; beta = 0.3 / RTT_max, as shown in lines 14 &amp; 15 of alpha = 0.1 * Tupdate / RTT_max^2;
beta = 0.3 / RTT_max,
as shown in lines 13 and 14 of
<xref target="dualq_fig_Algo_pi2_core_header" format="default"/>. These are derived <xref target="dualq_fig_Algo_pi2_core_header" format="default"/>. These are derived
from the stability analysis in <xref target="PI2" format="default"/>. Fo r the default from the stability analysis in <xref target="PI2" format="default"/>. Fo r the default
values of Tupdate=16 ms and RTT_max = 100 ms, they result in alpha = values of Tupdate = 16 ms and RTT_max = 100 ms, they result in alpha =
0.16; beta = 3.2 (discrepancies are due to rounding). These defaults 0.16; beta = 3.2 (discrepancies are due to rounding). These defaults
have been verified with a wide range of link rates, target delays and have been verified with a wide range of link rates, target delays, and
a range of traffic models with mixed and similar RTTs, short and long traffic models with mixed and similar RTTs, short and long
flows, etc.</t> flows, etc.</t>
<t>In corner cases, p' can overflow the range [0,1] so the resulting <t>In corner cases, p' can overflow the range [0,1] so the resulting
value of p' has to be bounded (omitted from the pseudocode). Then, as value of p' has to be bounded (omitted from the pseudocode). Then, as
already explained, the coupled and Classic probabilities are derived already explained, the coupled and Classic probabilities are derived
from the new p' in lines 4 and 5 of <xref target="dualq_fig_Algo_pi2_cor e" format="default"/> as p_CL = k*p' and p_C = p'^2.</t> from the new p' in lines 4 and 5 of <xref target="dualq_fig_Algo_pi2_cor e" format="default"/> as p_CL = k*p' and p_C = p'^2.</t>
<t>Because the coupled L4S marking probability (p_CL) is factored up <t>Because the coupled L4S marking probability (p_CL) is factored up
by k, the dynamic gain parameters alpha and beta are also inherently by k, the dynamic gain parameters alpha and beta are also inherently
factored up by k for the L4S queue. So, the effective gain factor for factored up by k for the L4S queue. So, the effective gain factor for
the L4S queue is k*alpha (with defaults alpha = 0.16 Hz and k=2, the L4S queue is k*alpha (with defaults alpha = 0.16 Hz and k = 2,
effective L4S alpha = 0.32 Hz).</t> effective L4S alpha = 0.32 Hz).</t>
<t>Unlike in PIE <xref target="RFC8033" format="default"/>, alpha and be ta do not <t>Unlike in PIE <xref target="RFC8033" format="default"/>, alpha and be ta do not
need to be tuned every Tupdate dependent on p'. Instead, in PI2, alpha need to be tuned every Tupdate dependent on p'. Instead, in PI2, alpha
and beta are independent of p' because the squaring applied to Classic and beta are independent of p' because the squaring applied to Classic
traffic tunes them inherently. This is explained in <xref target="PI2" f ormat="default"/>, which also explains why this more principled approach traffic tunes them inherently. This is explained in <xref target="PI2" f ormat="default"/>, which also explains why this more principled approach
removes the need for most of the heuristics that had to be added to removes the need for most of the heuristics that had to be added to
PIE.</t> PIE.</t>
<t>Nonetheless, an implementer might wish to add selected details to <t>Nonetheless, an implementer might wish to add selected details to
either AQM. For instance the Linux reference DualPI2 implementation either AQM. For instance, the Linux reference DualPI2 implementation
includes the following (not shown in the pseudocode above):</t> includes the following (not shown in the pseudocode above):</t>
<ul spacing="normal"> <ul spacing="normal">
<li>Classic and coupled marking or dropping (i.e. based on p_C <li>Classic and coupled marking or dropping (i.e., based on p_C
and p_CL from the PI controller) is not applied to a packet if the and p_CL from the PI controller) is not applied to a packet if the
aggregate queue length in bytes is &lt; 2 MTU (prior to enqueuing aggregate queue length in bytes is &lt; 2 MTU (prior to enqueuing
the packet or dequeuing it, depending on whether the AQM is the packet or dequeuing it, depending on whether the AQM is
configured to be applied at enqueue or dequeue);</li> configured to be applied at enqueue or dequeue); and</li>
<li>In the WRR scheduler, the 'credit' indicating which queue <li>in the WRR scheduler, the 'credit' indicating which queue
should transmit is only changed if there are packets in both should transmit is only changed if there are packets in both
queues (i.e. if there is actual resource contention). This queues (i.e., if there is actual resource contention). This
means that a properly paced L flow might never be delayed by the means that a properly paced L flow might never be delayed by the
WRR. The WRR credit is reset in favour of the L queue when the WRR. The WRR credit is reset in favour of the L queue when the
link is idle.</li> link is idle.</li>
</ul> </ul>
<t>An implementer might also wish to add other heuristics, <t>An implementer might also wish to add other heuristics,
e.g. burst protection <xref target="RFC8033" format="default"/> or enhan e.g., burst protection <xref target="RFC8033" format="default"/> or enha
ced nced
burst protection <xref target="RFC8034" format="default"/>.</t> burst protection <xref target="RFC8034" format="default"/>.</t>
<t>Notes:</t> <t>Notes:</t>
<ol spacing="normal" type="a"><li anchor="dualq_note_qdelay"> <ol spacing="normal" type="a">
<li anchor="note_qdelay">
<t>The drain rate of the queue can vary <t>The drain rate of the queue can vary
if it is scheduled relative to other queues, or to cater for if it is scheduled relative to other queues or if it accommodates
fluctuations in a wireless medium. To auto-adjust to changes in fluctuations in a wireless medium. To auto-adjust to changes in
drain rate, the queue needs to be measured in time, not bytes or drain rate, the queue needs to be measured in time, not bytes or
packets <xref target="AQMmetrics" format="default"/>, <xref target=" packets <xref target="AQMmetrics" format="default"/> <xref target="C
CoDel" format="default"/>. oDel" format="default"/>.
Queuing delay could be measured directly as the sojourn time (aka. Queuing delay could be measured directly as the sojourn time (a.k.a.
service time) of the queue, by storing a per-packet time-stamp as service time) of the queue by storing a per-packet timestamp as
each packet is enqueued, and subtracting this from the system time each packet is enqueued and subtracting it from the system time
when the packet is dequeued. If time-stamping is not easy to when the packet is dequeued. If timestamping is not easy to
introduce with certain hardware, queuing delay could be predicted introduce with certain hardware, queuing delay could be predicted
indirectly by dividing the size of the queue by the predicted indirectly by dividing the size of the queue by the predicted
departure rate, which might be known precisely for some link departure rate, which might be known precisely for some link
technologies (see for example in DOCSIS PIE [RFC8034]). </t> technologies (see, for example, DOCSIS PIE <xref target="RFC8034"/>) . </t>
<t>However, sojourn time is slow to detect bursts. <t>However, sojourn time is slow to detect bursts.
For instance, if a burst arrives at an empty queue, the sojourn For instance, if a burst arrives at an empty queue, the sojourn
time only fully measures the burst's delay when its last packet is time only fully measures the burst's delay when its last packet is
dequeued, even though the queue has known the size of the burst dequeued, even though the queue has known the size of the burst
since its last packet was enqueued - so it could have signalled since its last packet was enqueued -- so it could have signalled
congestion earlier. To remedy this, each head packet can be marked congestion earlier. To remedy this, each head packet can be marked
when it is dequeued based on the expected delay of the tail packet when it is dequeued based on the expected delay of the tail packet
behind it, as explained below, rather than based on the head behind it, as explained below, rather than based on the head
packet's own delay due to the packets in front of it. <xref target=" Heist21" format="default"/> identifies a specific scenario where bursty packet's own delay due to the packets in front of it. "Underutilizat ion with Bursty Traffic" in <xref target="Heist21" format="default"/> identifies a specific scenario where bursty
traffic significantly hits utilization of the L queue. If this traffic significantly hits utilization of the L queue. If this
effect proves to be more widely applicable, using the delay behind effect proves to be more widely applicable, using the delay behind
the head could improve performance.</t> the head could improve performance.</t>
<t>The <t>The
delay behind the head can be implemented by dividing the backlog delay behind the head can be implemented by dividing the backlog
at dequeue by the link rate or equivalently multiplying the at dequeue by the link rate or equivalently multiplying the
backlog by the delay per unit of backlog. The implementation backlog by the delay per unit of backlog. The implementation
details will depend on whether the link rate is known; if it is details will depend on whether the link rate is known; if it is
not, a moving average of the delay per unit backlog can be not, a moving average of the delay per unit backlog can be
maintained. This delay consists of serialization as well as media maintained. This delay consists of serialization as well as media
acquisition for shared media. So the details will depend strongly acquisition for shared media. So the details will depend strongly
on the specific link technology, This approach should be less on the specific link technology. This approach should be less
sensitive to timing errors and cost less in operations and memory sensitive to timing errors and cost less in operations and memory
than the otherwise equivalent 'scaled sojourn time' metric, which than the otherwise equivalent 'scaled sojourn time' metric, which
is the sojourn time of a packet scaled by the ratio of the queue is the sojourn time of a packet scaled by the ratio of the queue
sizes when the packet departed and arrived <xref target="SigQ-Dyn" f ormat="default"/>.</t> sizes when the packet departed and arrived <xref target="SigQ-Dyn" f ormat="default"/>.</t>
</li> </li>
<li>Line 2 of the dualpi2_enqueue() function (<xref target="dualq_fig_ Algo_pi2_enqueue" format="default"/>) assumes an implementation <li anchor="note_separate_buffers">Line 2 of the dualpi2_enqueue() fun ction (<xref target="dualq_fig_Algo_pi2_enqueue" format="default"/>) assumes an implementation
where lq and cq share common buffer memory. An alternative where lq and cq share common buffer memory. An alternative
implementation could use separate buffers for each queue, in which implementation could use separate buffers for each queue, in which
case the arriving packet would have to be classified first to case the arriving packet would have to be classified first to
determine which buffer to check for available space. The choice is determine which buffer to check for available space. The choice is
a trade-off; a shared buffer can use less memory whereas separate a trade-off; a shared buffer can use less memory whereas separate
buffers isolate the L4S queue from tail-drop due to large bursts buffers isolate the L4S queue from tail drop due to large bursts
of Classic traffic (e.g. a Classic Reno TCP during slow-start of Classic traffic (e.g., a Classic Reno TCP during slow-start
over a long RTT).</li> over a long RTT).</li>
<li> <li anchor="note_ramp">
<t>There has been some concern that using the step function of <t>There has been some concern that using the step function of
DCTCP for the Native L4S AQM requires end-systems to smooth the DCTCP for the Native L4S AQM requires end systems to smooth the
signal for an unnecessarily large number of round trips to ensure signal for an unnecessarily large number of round trips to ensure
sufficient fidelity. A ramp is no worse than a step in initial sufficient fidelity. A ramp is no worse than a step in initial
experiments with existing DCTCP. Therefore, it is recommended that experiments with existing DCTCP. Therefore, it is recommended that
a ramp is configured in place of a step, which will allow a ramp is configured in place of a step, which will allow
congestion control algorithms to investigate faster smoothing congestion control algorithms to investigate faster smoothing
algorithms.</t> algorithms.</t>
<t>A ramp is more general that a <t>A ramp is more general than a
step, because an operator can effectively turn the ramp into a step, because an operator can effectively turn the ramp into a
step function, as used by DCTCP, by setting the range to zero. step function, as used by DCTCP, by setting the range to zero.
There will not be a divide by zero problem at line 5 of <xref target ="dualq_fig_Algo_laqm_core" format="default"/> because, if minTh is equal to There will not be a divide by zero problem at line 5 of <xref target ="dualq_fig_Algo_laqm_core" format="default"/> because, if minTh is equal to
maxTh, the condition for this ramp calculation cannot arise.</t> maxTh, the condition for this ramp calculation cannot arise.</t>
</li> </li>
</ol> </ol>
</section> </section>
<section anchor="dualq_Ex_algo_pi2-2" numbered="true" toc="default"> <section anchor="dualq_Ex_algo_pi2-2" numbered="true" toc="default">
<name>Pass #2: Edge-Case Details</name> <name>Pass #2: Edge-Case Details</name>
<t>This section takes a second pass through the pseudocode adding <t>This section takes a second pass through the pseudocode to add
details of two edge-cases: low link rate and overload. <xref target="dua lq_fig_Algo_pi2_full_dequeue" format="default"/> repeats the dequeue details of two edge-cases: low link rate and overload. <xref target="dua lq_fig_Algo_pi2_full_dequeue" format="default"/> repeats the dequeue
function of <xref target="dualq_fig_Algo_pi2_dequeue" format="default"/> , but with function of <xref target="dualq_fig_Algo_pi2_dequeue" format="default"/> , but with
details of both edge-cases added. Similarly, <xref target="dualq_fig_Alg o_pi2_full_core" format="default"/> repeats the core PI algorithm details of both edge-cases added. Similarly, <xref target="dualq_fig_Alg o_pi2_full_core" format="default"/> repeats the core PI algorithm
of <xref target="dualq_fig_Algo_pi2_core" format="default"/>, but with o verload details of <xref target="dualq_fig_Algo_pi2_core" format="default"/>, but with o verload details
added. The initialization, enqueue, L4S AQM and recur functions are added. The initialization, enqueue, L4S AQM, and recur functions are
unchanged.</t> unchanged.</t>
<t>The link rate can be so low that it takes a single packet queue <t>The link rate can be so low that it takes a single packet queue
longer to serialize than the threshold delay at which ECN marking longer to serialize than the threshold delay at which ECN marking
starts to be applied in the L queue. Therefore, a minimum marking starts to be applied in the L queue. Therefore, a minimum marking
threshold parameter in units of packets rather than time is necessary threshold parameter in units of packets rather than time is necessary
(Th_len, default 1 packet in line 19 of <xref target="dualq_fig_Algo_pi2 _core_header" format="default"/>) to ensure that the ramp (Th_len, default 1 packet in line 19 of <xref target="dualq_fig_Algo_pi2 _core_header" format="default"/>) to ensure that the ramp
does not trigger excessive marking on slow links. Where an does not trigger excessive marking on slow links. Where an
implementation knows the link rate, it can set up this minimum at the implementation knows the link rate, it can set up this minimum at the
time it is configured. For instance, it would divide 1 MTU by the link time it is configured.
For instance, it would divide 1 MTU by the link
rate to convert it into a serialization time, then if the lower rate to convert it into a serialization time, then if the lower
threshold of the Native L AQM ramp was lower than this serialization threshold of the Native L AQM ramp was lower than this serialization
time, it could increase the thresholds to shift the bottom of the ramp time, it could increase the thresholds to shift the bottom of the ramp
to 2 MTU. This is the approach used in DOCSIS <xref target="DOCSIS3.1" f ormat="default"/>, because the configured link rate is dedicated to to 2 MTU. This is the approach used in DOCSIS <xref target="DOCSIS3.1" f ormat="default"/>, because the configured link rate is dedicated to
the DualQ.</t> the DualQ.</t>
<t>The pseudocode given here applies where the link rate is unknown, <t>The pseudocode given here applies where the link rate is unknown,
which is more common for software implementations that might be which is more common for software implementations that might be
deployed in scenarios where the link is shared with other queues. In deployed in scenarios where the link is shared with other queues. In
lines 5a to 5d in <xref target="dualq_fig_Algo_pi2_full_dequeue" format= "default"/> the lines 5a to 5d in <xref target="dualq_fig_Algo_pi2_full_dequeue" format= "default"/>, the
native L4S marking probability, p'_L, is zeroed if the queue is only 1 native L4S marking probability, p'_L, is zeroed if the queue is only 1
packet (in the default configuration).</t> packet (in the default configuration).</t>
<t>Linux implementation note:</t> <aside><t>Linux implementation note: In Linux, the check that the
<ul spacing="normal"> queue exceeds Th_len before marking with the Native L4S AQM is
<li>In Linux, the check that the queue exceeds Th_len before actually at enqueue, not dequeue; otherwise, it would exempt the last
marking with the native L4S AQM is actually at enqueue, not packet of a burst from being marked. The result of the check is
dequeue, otherwise it would exempt the last packet of a burst from conveyed from enqueue to the dequeue function via a boolean in the
being marked. The result of the check is conveyed from enqueue to packet metadata.</t>
the dequeue function via a boolean in the packet metadata.</li> </aside>
</ul>
<t>Persistent overload is deemed to have occurred when Classic <t>Persistent overload is deemed to have occurred when Classic
drop/marking probability reaches p_Cmax. Above this point, the Classic drop/marking probability reaches p_Cmax. Above this point, the Classic
drop probability is applied to both L and C queues, irrespective of drop probability is applied to both the L and C queues, irrespective of
whether any packet is ECN-capable. ECT packets that are not dropped whether any packet is ECN-capable. ECT packets that are not dropped
can still be ECN-marked.</t> can still be ECN-marked.</t>
<t>In line 10 of the initialization function (<xref target="dualq_fig_Al
go_pi2_core_header" format="default"/>), the maximum Classic drop <t>In line 11 of the initialization function (<xref target="dualq_fig_Al
go_pi2_core_header" format="default"/>), the maximum Classic drop
probability p_Cmax = min(1/k^2, 1) or 1/4 for the default coupling probability p_Cmax = min(1/k^2, 1) or 1/4 for the default coupling
factor k=2. In practice, 25% has been found to be a good threshold to factor k = 2. In practice, 25% has been found to be a good threshold to
preserve fairness between ECN capable and non ECN capable traffic. preserve fairness between ECN-capable and non-ECN-capable traffic.
This protects the queues against both temporary overload from This protects the queues against both temporary overload from
responsive flows and more persistent overload from any unresponsive responsive flows and more persistent overload from any unresponsive
traffic that falsely claims to be responsive to ECN.</t> traffic that falsely claims to be responsive to ECN.</t>
<t>When the Classic ECN marking probability reaches the p_Cmax <t>When the Classic ECN-marking probability reaches the p_Cmax
threshold (1/k^2), the marking probability coupled to the L4S queue, threshold (1/k^2), the marking probability that is coupled to the L4S qu
p_CL will always be 100% for any k (by equation (1) in <xref target="dua eue,
lq_algo" format="default"/>). So, for readability, the constant p_Lmax is p_CL, will always be 100% for any k (by equation (1) in <xref target="du
defined as 1 in line 22 of the initialization function (<xref target="du alq_coupled" format="default"/>). So, for readability, the constant p_Lmax is
alq_fig_Algo_pi2_core_header" format="default"/>). This is intended to ensure defined as 1 in line 21 of the initialization function (<xref target="du
that the L4S queue starts to introduce dropping once ECN-marking alq_fig_Algo_pi2_core_header" format="default"/>). This is intended to ensure
saturates at 100% and can rise no further. The 'Prague L4S' that the L4S queue starts to introduce dropping once ECN marking
requirements <xref target="I-D.ietf-tsvwg-ecn-l4s-id" format="default"/> saturates at 100% and can rise no further. The 'Prague L4S
state requirements' <xref target="RFC9331" format="default"/> state
that, when an L4S congestion control detects a drop, it falls back to that when an L4S congestion control detects a drop, it falls back to
a response that coexists with 'Classic' Reno congestion control. So it a response that coexists with 'Classic' Reno congestion control. So, it
is correct that, when the L4S queue drops packets, it drops them is correct that when the L4S queue drops packets, it drops them
proportional to p'^2, as if they are Classic packets.</t> proportional to p'^2, as if they are Classic packets.</t>
<t>The two queues each test for overload in lines 4b and 12b of the <t>The two queues each test for overload in lines 4b and 12b of the
dequeue function (<xref target="dualq_fig_Algo_pi2_full_dequeue" format= "default"/>). dequeue function (<xref target="dualq_fig_Algo_pi2_full_dequeue" format= "default"/>).
Lines 8c to 8g drop L4S packets with probability p'^2. Lines 8h to 8i Lines 8c to 8g drop L4S packets with probability p'^2. Lines 8h to 8i
mark the remaining packets with probability p_CL. Given p_Lmax = 1, mark the remaining packets with probability p_CL. Given p_Lmax = 1,
all remaining packets will be marked because, to have reached the else all remaining packets will be marked because, to have reached the else
block at line 8b, p_CL &gt;= 1.</t> block at line 8b, p_CL &gt;= 1.</t>
<t>Line 2a in the core PI algorithm (<xref target="dualq_fig_Algo_pi2_fu ll_core" format="default"/>) deals with overload of the <t>Line 2a in the core PI algorithm (<xref target="dualq_fig_Algo_pi2_fu ll_core" format="default"/>) deals with overload of the
L4S queue when there is little or no Classic traffic. This is L4S queue when there is little or no Classic traffic. This is
necessary, because the core PI algorithm maintains the appropriate necessary, because the core PI algorithm maintains the appropriate
drop probability to regulate overload, but it depends on the length of drop probability to regulate overload, but it depends on the length of
the Classic queue. If there is little or no Classic queue the naive PI the Classic queue. If there is little or no Classic queue, the naive PI-
update function in <xref target="dualq_fig_Algo_pi2_core" format="defaul update function
t"/> would drop (<xref target="dualq_fig_Algo_pi2_core" format="default"/>) would drop
nothing, even if the L4S queue were overloaded - so tail drop would nothing, even if the L4S queue were overloaded -- so tail drop would
have to take over (lines 2 and 3 of <xref target="dualq_fig_Algo_pi2_enq ueue" format="default"/>).</t> have to take over (lines 2 and 3 of <xref target="dualq_fig_Algo_pi2_enq ueue" format="default"/>).</t>
<t>Instead, line 2a of the full PI update function in <xref target="dual q_fig_Algo_pi2_full_core" format="default"/> ensures that the base PI AQM <t>Instead, line 2a of the full PI-update function (<xref target="dualq_ fig_Algo_pi2_full_core" format="default"/>) ensures that the Base PI AQM
in line 3 is driven by whichever of the two queue delays is greater, in line 3 is driven by whichever of the two queue delays is greater,
but line 3 still always uses the same Classic target (default 15 ms). but line 3 still always uses the same Classic target (default 15 ms).
If L queue delay is greater just because there is little or no Classic If L queue delay is greater just because there is little or no Classic
traffic, normally it will still be well below the base AQM target. traffic, normally it will still be well below the Base AQM target.
This is because L4S traffic is also governed by the shallow threshold This is because L4S traffic is also governed by the shallow threshold
of its own native AQM (lines 5 and 6 of the dequeue algorithm in <xref t of its own Native AQM (lines 5a to 6 of the dequeue algorithm in <xref t
arget="dualq_fig_Algo_pi2_full_dequeue" format="default"/>). So the base AQM wil arget="dualq_fig_Algo_pi2_full_dequeue" format="default"/>). So the Base AQM wil
l be l be
driven to zero and not contribute. However, if the L queue is driven to zero and not contribute.
However, if the L queue is
overloaded by traffic that is unresponsive to its marking, the max() overloaded by traffic that is unresponsive to its marking, the max()
in line 2 enables the L queue to smoothly take over driving the base in line 2a of <xref target="dualq_fig_Algo_pi2_full_core" format="defaul t"/> enables the L queue to smoothly take over driving the Base
AQM into overload mode even if there is little or no Classic traffic. AQM into overload mode even if there is little or no Classic traffic.
Then the base AQM will keep the L queue to the Classic target (default Then the Base AQM will keep the L queue to the Classic target (default
15 ms) by shedding L packets.</t> 15 ms) by shedding L packets.</t>
<figure anchor="dualq_fig_Algo_pi2_full_dequeue"> <figure anchor="dualq_fig_Algo_pi2_full_dequeue">
<name>Example Dequeue Pseudocode for DualQ Coupled PI2 AQM (Including Code for Edge-Cases)</name> <name>Example Dequeue Pseudocode for DualQ Coupled PI2 AQM (Including Code for Edge-Cases)</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: dualpi2_dequ <sourcecode><![CDATA[
eue(lq, cq, pkt) { % Couples L4S & Classic queues 1: dualpi2_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: while ( lq.byt() + cq.byt() > 0 ) { 2: while ( lq.byt() + cq.byt() > 0 ) {
3: if ( scheduler() == lq ) { 3: if ( scheduler() == lq ) {
4a: lq.dequeue(pkt) % L4S scheduled 4a: lq.dequeue(pkt) % L4S scheduled
4b: if ( p_CL < p_Lmax ) { % Check for overload saturation 4b: if ( p_CL < p_Lmax ) { % Check for overload saturation
5a: if (lq.len()>Th_len) % >1 packet queued 5a: if (lq.len()>Th_len) % >1 packet queued
5b: p'_L = laqm(lq.time()) % Native LAQM 5b: p'_L = laqm(lq.time()) % Native LAQM
5c: else 5c: else
5d: p'_L = 0 % Suppress marking 1 pkt queue 5d: p'_L = 0 % Suppress marking 1 pkt queue
6: p_L = max(p'_L, p_CL) % Combining function 6: p_L = max(p'_L, p_CL) % Combining function
7: if ( recur(lq, p_L) %Linear marking 7: if ( recur(lq, p_L) %Linear marking
skipping to change at line 3075 skipping to change at line 2459
13: drop(pkt) % squared drop, redo loop 13: drop(pkt) % squared drop, redo loop
14: continue % continue to the top of the while loop 14: continue % continue to the top of the while loop
15: } 15: }
16: mark(pkt) % squared mark 16: mark(pkt) % squared mark
17: } 17: }
18: } 18: }
19: return(pkt) % return the packet and stop 19: return(pkt) % return the packet and stop
20: } 20: }
21: return(NULL) % no packet to dequeue 21: return(NULL) % no packet to dequeue
22: } 22: }
]]></artwork> ]]></sourcecode>
</figure> </figure>
<figure anchor="dualq_fig_Algo_pi2_full_core"> <figure anchor="dualq_fig_Algo_pi2_full_core">
<name>Example PI-Update Pseudocode for DualQ Coupled PI2 AQM (Includin <name>Example PI-update Pseudocode for DualQ Coupled PI2 AQM (Includin
g Overload Code)</name> g Overload Code)</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: dualpi2_upda <sourcecode><![CDATA[
te(lq, cq) { % Update p' every Tupdate 1: dualpi2_update(lq, cq) { % Update p' every Tupdate
2a: curq = max(cq.time(), lq.time()) % use greatest queuing time 2a: curq = max(cq.time(), lq.time()) % use greatest queuing time
3: p' = p' + alpha * (curq - target) + beta * (curq - prevq) 3: p' = p' + alpha * (curq - target) + beta * (curq - prevq)
4: p_CL = p' * k % Coupled L4S prob = base prob * coupling factor 4: p_CL = p' * k % Coupled L4S prob = base prob * coupling factor
5: p_C = p'^2 % Classic prob = (base prob)^2 5: p_C = p'^2 % Classic prob = (base prob)^2
6: prevq = curq 6: prevq = curq
7: } 7: }
]]></artwork> ]]></sourcecode>
</figure> </figure>
<t/> <t/>
<t>The choice of scheduler technology is critical to overload <t>The choice of scheduler technology is critical to overload
protection (see <xref target="dualq_Overload_Starvation" format="default "/>). </t> protection (see <xref target="dualq_Overload_Starvation" format="default "/>). </t>
<ul spacing="normal"> <ul spacing="normal">
<li>A well-understood weighted scheduler such as weighted <li>A well-understood weighted scheduler such as WRR is recommended. A
round-robin (WRR) is recommended. As long as the scheduler weight s long as the scheduler weight
for Classic is small (e.g. 1/16), its exact value is for Classic is small (e.g., 1/16), its exact value is
unimportant because it does not normally determine capacity unimportant, because it does not normally determine capacity
shares. The weight is only important to prevent unresponsive L4S shares. The weight is only important to prevent unresponsive L4S
traffic starving Classic traffic in the short term (see <xref target ="dualq_Overload_Starvation" format="default"/>). This is because capacity traffic starving Classic traffic in the short term (see <xref target ="dualq_Overload_Starvation" format="default"/>). This is because capacity
sharing between the queues is normally determined by the coupled sharing between the queues is normally determined by the coupled
congestion signal, which overrides the scheduler, by making L4S congestion signal, which overrides the scheduler, by making L4S
sources leave roughly equal per-flow capacity available for sources leave roughly equal per-flow capacity available for
Classic flows.</li> Classic flows.</li>
<li> <li>
<t>Alternatively, a time-shifted FIFO (TS-FIFO) could be used. It <t>Alternatively, a time-shifted FIFO (TS-FIFO) could be used. It
works by selecting the head packet that has waited the longest, works by selecting the head packet that has waited the longest,
biased against the Classic traffic by a time-shift of tshift. To biased against the Classic traffic by a time-shift of tshift. To
implement time-shifted FIFO, the scheduler() function in line 3 of implement TS-FIFO, the scheduler() function in line 3 of
the dequeue code would simply be implemented as the scheduler() the dequeue code would simply be implemented as the scheduler()
function at the bottom of <xref target="dualq_fig_Algo_Real" format= "default"/> in function at the bottom of <xref target="dualq_fig_Algo_Real" format= "default"/> in
<xref target="dualq_Ex_algo" format="default"/>. For the public Inte <xref target="dualq_Ex_algo" format="default"/>. For the public Inte
rnet a good rnet, a good
value for tshift is 50ms. For private networks with smaller value for tshift is 50 ms. For private networks with smaller
diameter, about 4*target would be reasonable. TS-FIFO is a very diameter, about 4*target would be reasonable. TS-FIFO is a very
simple scheduler, but complexity might need to be added to address simple scheduler, but complexity might need to be added to address
some deficiencies (which is why it is not recommended over some deficiencies (which is why it is not recommended over
WRR):</t> WRR):</t>
<ul spacing="normal"> <ul spacing="normal">
<li>TS-FIFO does not fully isolate latency in the L4S queue <li>TS-FIFO does not fully isolate latency in the L4S queue
from uncontrolled bursts in the Classic queue;</li> from uncontrolled bursts in the Classic queue;</li>
<li>Using sojourn time for TS-FIFO is only appropriate if <li>using sojourn time for TS-FIFO is only appropriate if
time-stamping of packets is feasible;</li> timestamping of packets is feasible; and</li>
<li>Even if time-stamping is supported, the sojourn time of the <li>even if timestamping is supported, the sojourn time of the
head packet is always stale, so a more instantaneous measure head packet is always stale, so a more instantaneous measure
of queue delay could be used (see Note a in <xref target="dualq_ Ex_algo_pi2-1" format="default"/>).</li> of queue delay could be used (see <xref target="note_qdelay" for mat="none">Note a</xref> in <xref target="dualq_Ex_algo_pi2-1" format="default"/ >).</li>
</ul> </ul>
</li> </li>
<li>A strict priority scheduler would be inappropriate as discussed <li>A strict priority scheduler would be inappropriate as discussed
in <xref target="dualq_Overload_Starvation" format="default"/>.</li> in <xref target="dualq_Overload_Starvation" format="default"/>.</li>
</ul> </ul>
</section> </section>
</section> </section>
<section anchor="dualq_Ex_algo" numbered="true" toc="default"> <section anchor="dualq_Ex_algo" numbered="true" toc="default">
<name>Example DualQ Coupled Curvy RED Algorithm</name> <name>Example DualQ Coupled Curvy RED Algorithm</name>
<t>As another example of a DualQ Coupled AQM algorithm, the pseudocode <t>As another example of a DualQ Coupled AQM algorithm, the pseudocode
below gives the Curvy RED based algorithm. Although the AQM was designed below gives the Curvy-RED-based algorithm. Although the AQM was designed
to be efficient in integer arithmetic, to aid understanding it is first to be efficient in integer arithmetic, to aid understanding it is first
given using floating point arithmetic (<xref target="dualq_fig_Algo_Real" format="default"/>). Then, one possible optimization for given using floating point arithmetic (<xref target="dualq_fig_Algo_Real" format="default"/>). Then, one possible optimization for
integer arithmetic is given, also in pseudocode (<xref target="dualq_fig_A lgo_Int" format="default"/>). To aid comparison, the line numbers are integer arithmetic is given, also in pseudocode (<xref target="dualq_fig_A lgo_Int" format="default"/>). To aid comparison, the line numbers are
kept in step between the two by using letter suffixes where the longer kept in step between the two by using letter suffixes where the longer
code needs extra lines.</t> code needs extra lines.</t>
<section anchor="dualq_Ex_algo_float" numbered="true" toc="default"> <section anchor="dualq_Ex_algo_float" numbered="true" toc="default">
<name>Curvy RED in Pseudocode</name> <name>Curvy RED in Pseudocode</name>
<t>The pseudocode manipulates three main structures of variables: the <t>The pseudocode manipulates three main structures of variables: the
packet (pkt), the L4S queue (lq) and the Classic queue (cq) and packet (pkt), the L4S queue (lq), and the Classic queue (cq). It is defi
consists of the following five functions:</t> ned
and described below in the following three functions:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>The initialization function cred_params_init(...) (<xref target="d ualq_fig_Algo_pi2_core_header" format="default"/>) that sets parameter <li>the initialization function cred_params_init(...) (<xref target="d ualq_fig_Algo_pi2_core_header" format="default"/>) that sets parameter
defaults (the API for setting non-default values is omitted for defaults (the API for setting non-default values is omitted for
brevity);</li> brevity);</li>
<li>The dequeue function cred_dequeue(lq, cq, pkt) (<xref target="dual <li>the dequeue function cred_dequeue(lq, cq, pkt) (<xref target="dual
q_fig_Algo_pi2_dequeue" format="default"/>);</li> q_fig_Algo_pi2_dequeue" format="default"/>); and</li>
<li>The scheduling function scheduler(), which selects between the <li>the scheduling function scheduler(), which selects between the
head packets of the two queues.</li> head packets of the two queues.</li>
</ul> </ul>
<t>It also uses the following functions that are either shown <t>It also uses the following functions that are either shown
elsewhere, or not shown in full here:</t> elsewhere or not shown in full here:</t>
<ul spacing="normal"> <ul spacing="normal">
<li>The enqueue function, which is identical to that used for <li>the enqueue function, which is identical to that used for
DualPI2, dualpi2_enqueue(lq, cq, pkt) in <xref target="dualq_fig_Alg o_pi2_enqueue" format="default"/>;</li> DualPI2, dualpi2_enqueue(lq, cq, pkt) in <xref target="dualq_fig_Alg o_pi2_enqueue" format="default"/>;</li>
<li>mark(pkt) and drop(pkt) for ECN-marking and dropping a <li>mark(pkt) and drop(pkt) for ECN marking and dropping a
packet;</li> packet;</li>
<li>cq.byt() or lq.byt() returns the current length <li>cq.byt() or lq.byt() returns the current length
(aka. backlog) of the relevant queue in bytes;</li> (a.k.a. backlog) of the relevant queue in bytes; and</li>
<li>cq.time() or lq.time() returns the current queuing delay of the <li>cq.time() or lq.time() returns the current queuing delay of the
relevant queue in units of time (see Note a in <xref target="dualq_E x_algo_pi2-1" format="default"/>).</li> relevant queue in units of time (see <xref target="note_qdelay" form at="none">Note a</xref> in <xref target="dualq_Ex_algo_pi2-1" format="default"/> ).</li>
</ul> </ul>
<t>Because Curvy RED was evaluated before DualPI2, certain <t>Because Curvy RED was evaluated before DualPI2, certain
improvements introduced for DualPI2 were not evaluated for Curvy RED. improvements introduced for DualPI2 were not evaluated for Curvy RED.
In the pseudocode below, the straightforward improvements have been In the pseudocode below, the straightforward improvements have been
added on the assumption they will provide similar benefits, but that added on the assumption they will provide similar benefits, but that
has not been proven experimentally. They are: i) a conditional has not been proven experimentally. They are: i) a conditional
priority scheduler instead of strict priority ii) a time-based priority scheduler instead of strict priority; ii) a time-based
threshold for the native L4S AQM; iii) ECN support for the Classic threshold for the Native L4S AQM; and iii) ECN support for the Classic
AQM. A recent evaluation has proved that a minimum ECN-marking AQM. A recent evaluation has proved that a minimum ECN-marking
threshold (minTh) greatly improves performance, so this is also threshold (minTh) greatly improves performance, so this is also
included in the pseudocode.</t> included in the pseudocode.</t>
<t>Overload protection has not been added to the Curvy RED pseudocode <t>Overload protection has not been added to the Curvy RED pseudocode
below so as not to detract from the main features. It would be added below so as not to detract from the main features. It would be added
in exactly the same way as in <xref target="dualq_Ex_algo_pi2-2" format= "default"/> for in exactly the same way as in <xref target="dualq_Ex_algo_pi2-2" format= "default"/> for
the DualPI2 pseudocode. The native L4S AQM uses a step threshold, but the DualPI2 pseudocode. The Native L4S AQM uses a step threshold, but
a ramp like that described for DualPI2 could be used instead. The a ramp like that described for DualPI2 could be used instead. The
scheduler uses the simple TS-FIFO algorithm, but it could be replaced scheduler uses the simple TS-FIFO algorithm, but it could be replaced
with WRR.</t> with WRR.</t>
<t>The Curvy RED algorithm has not been maintained or evaluated to the <t>The Curvy RED algorithm has not been maintained or evaluated to the
same degree as the DualPI2 algorithm. In initial experiments on same degree as the DualPI2 algorithm. In initial experiments on
broadband access links ranging from 4 Mb/s to 200 Mb/s with base RTTs broadband access links ranging from 4 Mb/s to 200 Mb/s with base RTTs
from 5 ms to 100 ms, Curvy RED achieved good results with the default from 5 ms to 100 ms, Curvy RED achieved good results with the default
parameters in <xref target="dualq_fig_Algo_cred_core_header" format="def ault"/>.</t> parameters in <xref target="dualq_fig_Algo_cred_core_header" format="def ault"/>.</t>
<t>The parameters are categorised by whether they relate to the <t>The parameters are categorized by whether they relate to the
Classic AQM, the L4S AQM or the framework coupling them together. Classic AQM, the L4S AQM, or the framework coupling them together.
Constants and variables derived from these parameters are also Constants and variables derived from these parameters are also
included at the end of each category. These are the raw input included at the end of each category. These are the raw input
parameters for the algorithm. A configuration front-end could accept parameters for the algorithm. A configuration front-end could accept
more meaningful parameters (e.g. RTT_max and RTT_typ) and convert more meaningful parameters (e.g., RTT_max and RTT_typ) and convert
them into these raw parameters, as has been done for DualPI2 in <xref ta rget="dualq_Ex_algo_pi2" format="default"/>. Where necessary, parameters are them into these raw parameters, as has been done for DualPI2 in <xref ta rget="dualq_Ex_algo_pi2" format="default"/>. Where necessary, parameters are
explained further in the walk-through of the pseudocode below.</t> explained further in the walk-through of the pseudocode below.</t>
<figure anchor="dualq_fig_Algo_cred_core_header"> <figure anchor="dualq_fig_Algo_cred_core_header">
<name>Example Header Pseudocode for DualQ Coupled Curvy RED AQM</name> <name>Example Header Pseudocode for DualQ Coupled Curvy RED AQM</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: cred_params_ <sourcecode><![CDATA[
init(...) { % Set input parameter defaults 1: cred_params_init(...) { % Set input parameter defaults
2: % DualQ Coupled framework parameters 2: % DualQ Coupled framework parameters
3: limit = MAX_LINK_RATE * 250 ms % Dual buffer size 3: limit = MAX_LINK_RATE * 250 ms % Dual buffer size
4: k' = 1 % Coupling factor as a power of 2 4: k' = 1 % Coupling factor as a power of 2
5: tshift = 50 ms % Time shift of TS-FIFO scheduler 5: tshift = 50 ms % Time-shift of TS-FIFO scheduler
6: % Constants derived from Classic AQM parameters 6: % Constants derived from Classic AQM parameters
7: k = 2^k' % Coupling factor from Equation (1) 7: k = 2^k' % Coupling factor from equation (1)
6: 6:
7: % Classic AQM parameters 7: % Classic AQM parameters
8: g_C = 5 % EWMA smoothing parameter as a power of 1/2 8: g_C = 5 % EWMA smoothing parameter as a power of 1/2
9: S_C = -1 % Classic ramp scaling factor as a power of 2 9: S_C = -1 % Classic ramp scaling factor as a power of 2
10: minTh = 500 ms % No Classic drop/mark below this queue delay 10: minTh = 500 ms % No Classic drop/mark below this queue delay
11: % Constants derived from Classic AQM parameters 11: % Constants derived from Classic AQM parameters
12: gamma = 2^(-g_C) % EWMA smoothing parameter 12: gamma = 2^(-g_C) % EWMA smoothing parameter
13: range_C = 2^S_C % Range of Classic ramp 13: range_C = 2^S_C % Range of Classic ramp
14: 14:
15: % L4S AQM parameters 15: % L4S AQM parameters
16: T = 1 ms % Queue delay threshold for native L4S AQM 16: T = 1 ms % Queue delay threshold for Native L4S AQM
17: % Constants derived from above parameters 17: % Constants derived from above parameters
18: S_L = S_C - k' % L4S ramp scaling factor as a power of 2 18: S_L = S_C - k' % L4S ramp scaling factor as a power of 2
19: range_L = 2^S_L % Range of L4S ramp 19: range_L = 2^S_L % Range of L4S ramp
20: } 20: }
]]></artwork> ]]></sourcecode>
</figure> </figure>
<figure anchor="dualq_fig_Algo_Real"> <figure anchor="dualq_fig_Algo_Real">
<name>Example Dequeue Pseudocode for DualQ Coupled Curvy RED AQM</name > <name>Example Dequeue Pseudocode for DualQ Coupled Curvy RED AQM</name >
<artwork name="" type="" align="left" alt=""><![CDATA[1: cred_dequeue <sourcecode><![CDATA[
(lq, cq, pkt) { % Couples L4S & Classic queues 1: cred_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: while ( lq.byt() + cq.byt() > 0 ) { 2: while ( lq.byt() + cq.byt() > 0 ) {
3: if ( scheduler() == lq ) { 3: if ( scheduler() == lq ) {
4: lq.dequeue(pkt) % L4S scheduled 4: lq.dequeue(pkt) % L4S scheduled
5a: p_CL = (Q_C - minTh) / range_L 5a: p_CL = (Q_C - minTh) / range_L
5b: if ( ( lq.time() > T ) 5b: if ( ( lq.time() > T )
5c: OR ( p_CL > maxrand(U) ) ) 5c: OR ( p_CL > maxrand(U) ) )
6: mark(pkt) 6: mark(pkt)
7: } else { 7: } else {
8: cq.dequeue(pkt) % Classic scheduled 8: cq.dequeue(pkt) % Classic scheduled
9a: Q_C = gamma * cq.time() + (1-gamma) * Q_C % Classic Q EWMA 9a: Q_C = gamma * cq.time() + (1-gamma) * Q_C % Classic Q EWMA
skipping to change at line 3260 skipping to change at line 2647
25: maxr = max(maxr, rand()) % 0 <= rand() < 1 25: maxr = max(maxr, rand()) % 0 <= rand() < 1
26: return(maxr) 26: return(maxr)
27: } 27: }
28: scheduler() { 28: scheduler() {
29: if ( lq.time() + tshift >= cq.time() ) 29: if ( lq.time() + tshift >= cq.time() )
30: return lq; 30: return lq;
31: else 31: else
32: return cq; 32: return cq;
33: } 33: }
]]></artwork> ]]></sourcecode>
</figure> </figure>
<t>The dequeue pseudocode (<xref target="dualq_fig_Algo_Real" format="de fault"/>) is <t>The dequeue pseudocode (<xref target="dualq_fig_Algo_Real" format="de fault"/>) is
repeatedly called whenever the lower layer is ready to forward a repeatedly called whenever the lower layer is ready to forward a
packet. It schedules one packet for dequeuing (or zero if the queue is packet. It schedules one packet for dequeuing (or zero if the queue is
empty) then returns control to the caller, so that it does not block empty) then returns control to the caller so that it does not block
while that packet is being forwarded. While making this dequeue while that packet is being forwarded. While making this dequeue
decision, it also makes the necessary AQM decisions on dropping or decision, it also makes the necessary AQM decisions on dropping or
marking. The alternative of applying the AQMs at enqueue would shift marking. The alternative of applying the AQMs at enqueue would shift
some processing from the critical time when each packet is dequeued. some processing from the critical time when each packet is dequeued.
However, it would also add a whole queue of delay to the control However, it would also add a whole queue of delay to the control
signals, making the control loop very sloppy.</t> signals, making the control loop very sloppy.</t>
<t>The code is written assuming the AQMs are applied on dequeue (Note <t>The code is written assuming the AQMs are applied on dequeue
<xref format="counter" target="dualq_note_dequeue"/>). All the dequeue (<xref format="none" target="dualq_note_dequeue">Note 1</xref>). All the
dequeue
code is contained within a large while loop so that if it decides to code is contained within a large while loop so that if it decides to
drop a packet, it will continue until it selects a packet to schedule. drop a packet, it will continue until it selects a packet to schedule.
If both queues are empty, the routine returns NULL at line 20. Line 3 If both queues are empty, the routine returns NULL at line 20. Line 3
of the dequeue pseudocode is where the conditional priority scheduler of the dequeue pseudocode is where the conditional priority scheduler
chooses between the L4S queue (lq) and the Classic queue (cq). The chooses between the L4S queue (lq) and the Classic queue (cq). The
time-shifted FIFO scheduler is shown at lines 28-33, which would be TS-FIFO scheduler is shown at lines 28-33, which would be
suitable if simplicity is paramount (see Note <xref format="counter" tar suitable if simplicity is paramount (see <xref format="none" target="dua
get="dualq_note_conditional_priority"/>).</t> lq_note_conditional_priority">Note 2</xref>).</t>
<t>Within each queue, the decision whether to forward, drop or mark is <t>Within each queue, the decision whether to forward, drop, or mark is
taken as follows (to simplify the explanation, it is assumed that taken as follows (to simplify the explanation, it is assumed that
U=1):</t> U = 1):</t>
<dl newline="false" spacing="normal"> <dl newline="true" spacing="normal">
<dt>L4S:</dt> <dt>L4S:</dt>
<dd> <dd>
<t>If the test at line 3 determines there is an <t>If the test at line 3 determines there is an
L4S packet to dequeue, the tests at lines 5b and 5c determine L4S packet to dequeue, the tests at lines 5b and 5c determine
whether to mark it. The first is a simple test of whether the L4S whether to mark it. The first is a simple test of whether the L4S
queue delay (lq.time()) is greater than a step threshold T (Note queue delay (lq.time()) is greater than a step threshold T
<xref format="counter" target="dualq_note_step"/>). The second (<xref target="dualq_note_step" format="none">Note 3</xref>). The se
test is similar to the random ECN marking in RED, but with the cond
test is similar to the random ECN marking in RED but with the
following differences: i) marking depends on queuing time, not following differences: i) marking depends on queuing time, not
bytes, in order to scale for any link rate without being bytes, in order to scale for any link rate without being
reconfigured; ii) marking of the L4S queue depends on a logical OR reconfigured; ii) marking of the L4S queue depends on a logical OR
of two tests; one against its own queuing time and one against the of two tests: one against its own queuing time and one against the
queuing time of the <em>other</em> (Classic) queuing time of the <em>other</em> (Classic)
queue; iii) the tests are against the instantaneous queuing time queue; iii) the tests are against the instantaneous queuing time
of the L4S queue, but a smoothed average of the other (Classic) of the L4S queue but against a smoothed average of the other (Classi
queue; iv) the queue is compared with the maximum of U random c)
numbers (but if U=1, this is the same as the single random number queue; and iv) the queue is compared with the maximum of U random
numbers (but if U = 1, this is the same as the single random number
used in RED).</t> used in RED).</t>
<t>Specifically, in line 5a the <t>Specifically, in line 5a, the
coupled marking probability p_CL is set to the amount by which the coupled marking probability p_CL is set to the amount by which the
averaged Classic queueing delay Q_C exceeds the minimum queuing averaged Classic queuing delay Q_C exceeds the minimum queuing
delay threshold (minTh) all divided by the L4S scaling parameter delay threshold (minTh), all divided by the L4S scaling parameter
range_L. range_L represents the queuing delay (in seconds) added range_L. range_L represents the queuing delay (in seconds) added
to minTh at which marking probability would hit 100%. Then in line to minTh at which marking probability would hit 100%. Then, in line
5c (if U=1) the result is compared with a uniformly distributed 5c (if U = 1), the result is compared with a uniformly distributed
random number between 0 and 1, which ensures that, over range_L, random number between 0 and 1, which ensures that, over range_L,
marking probability will linearly increase with queueing time.</t> marking probability will linearly increase with queuing time.</t>
</dd> </dd>
<dt>Classic:</dt> <dt>Classic:</dt>
<dd> <dd>
<t>If the scheduler at line 3 chooses to <t>If the scheduler at line 3 chooses to
dequeue a Classic packet and jumps to line 7, the test at line 10b dequeue a Classic packet and jumps to line 7, the test at line 10b
determines whether to drop or mark it. But before that, line 9a determines whether to drop or mark it. But before that, line 9a
updates Q_C, which is an exponentially weighted moving average updates Q_C, which is an exponentially weighted moving average
(Note <xref format="counter" target="dualq_note_non-EWMA"/>) of (Note <xref format="counter" target="dualq_note_non-EWMA"/>) of
the queuing time of the Classic queue, where cq.time() is the the queuing time of the Classic queue, where cq.time() is the
current instantaneous queueing time of the packet at the head of current instantaneous queuing time of the packet at the head of
the Classic queue (zero if empty) and gamma is the EWMA constant the Classic queue (zero if empty), and gamma is the exponentially we
(default 1/32, see line 12 of the initialization function). ighted moving average (EWMA) constant
(default 1/32; see line 12 of the initialization function).
</t> </t>
<t>Lines 10a and 10b implement the Classic <t>Lines 10a and 10b implement the Classic
AQM. In line 10a the averaged queuing time Q_C is divided by the AQM. In line 10a, the averaged queuing time Q_C is divided by the
Classic scaling parameter range_C, in the same way that queuing Classic scaling parameter range_C, in the same way that queuing
time was scaled for L4S marking. This scaled queuing time will be time was scaled for L4S marking. This scaled queuing time will be
squared to compute Classic drop probability so, before it is squared to compute Classic drop probability. So, before it is
squared, it is effectively the square root of the drop squared, it is effectively the square root of the drop
probability, hence it is given the variable name sqrt_p_C. The probability; hence, it is given the variable name sqrt_p_C. The
squaring is done by comparing it with the maximum out of two squaring is done by comparing it with the maximum out of two
random numbers (assuming U=1). Comparing it with the maximum out random numbers (assuming U = 1). Comparing it with the maximum out
of two is the same as the logical `AND' of two tests, which of two is the same as the logical 'AND' of two tests, which
ensures drop probability rises with the square of queuing ensures drop probability rises with the square of queuing
time.</t> time.</t>
</dd> </dd>
</dl> </dl>
<t>The AQM functions in each queue (lines 5c &amp; 10b) are two cases <t>The AQM functions in each queue (lines 5c and 10b) are two cases
of a new generalization of RED called Curvy RED, motivated as follows. of a new generalization of RED called 'Curvy RED', motivated as follows.
When the performance of this AQM was compared with FQ-CoDel and PIE, When the performance of this AQM was compared with FQ-CoDel and PIE,
their goal of holding queuing delay to a fixed target seemed their goal of holding queuing delay to a fixed target seemed
misguided <xref target="CRED_Insights" format="default"/>. As the number of flows misguided <xref target="CRED_Insights" format="default"/>. As the number of flows
increases, if the AQM does not allow host congestion controllers to increases, if the AQM does not allow host congestion controllers to
increase queuing delay, it has to introduce abnormally high levels of increase queuing delay, it has to introduce abnormally high levels of
loss. Then loss rather than queuing becomes the dominant cause of loss. Then loss rather than queuing becomes the dominant cause of
delay for short flows, due to timeouts and tail losses.</t> delay for short flows, due to timeouts and tail losses.</t>
<t>Curvy RED constrains delay with a softened target that allows some <t>Curvy RED constrains delay with a softened target that allows some
increase in delay as load increases. This is achieved by increasing increase in delay as load increases. This is achieved by increasing
drop probability on a convex curve relative to queue growth (the drop probability on a convex curve relative to queue growth (the
square curve in the Classic queue, if U=1). Like RED, the curve hugs square curve in the Classic queue, if U = 1). Like RED, the curve hugs
the zero axis while the queue is shallow. Then, as load increases, it the zero axis while the queue is shallow. Then, as load increases, it
introduces a growing barrier to higher delay. But, unlike RED, it introduces a growing barrier to higher delay. But, unlike RED, it
requires only two parameters, not three. The disadvantage of Curvy RED requires only two parameters, not three. The disadvantage of Curvy RED
(compared to a PI controller for example) is that it is not adapted to (compared to a PI controller, for example) is that it is not adapted to
a wide range of RTTs. Curvy RED can be used as is when the RTT range a wide range of RTTs. Curvy RED can be used as is when the RTT range
to be supported is limited, otherwise an adaptation mechanism is to be supported is limited; otherwise, an adaptation mechanism is
needed.</t> needed.</t>
<t>From our limited experiments with Curvy RED so far, recommended <t>From our limited experiments with Curvy RED so far, recommended
values of these parameters are: S_C = -1; g_C = 5; T = 5 * MTU at the values of these parameters are: S_C = -1; g_C = 5; T = 5 * MTU at the
link rate (about 1ms at 60Mb/s) for the range of base RTTs typical on link rate (about 1 ms at 60 Mb/s) for the range of base RTTs typical on
the public Internet. <xref target="CRED_Insights" format="default"/> exp lains why these the public Internet. <xref target="CRED_Insights" format="default"/> exp lains why these
parameters are applicable whatever rate link this AQM implementation parameters are applicable whatever rate link this AQM implementation
is deployed on and how the parameters would need to be adjusted for a is deployed on and how the parameters would need to be adjusted for a
scenario with a different range of RTTs (e.g. a data centre). The scenario with a different range of RTTs (e.g., a data centre). The
setting of k depends on policy (see <xref target="dualq_norm_reqs" forma t="default"/> setting of k depends on policy (see <xref target="dualq_norm_reqs" forma t="default"/>
and <xref target="dualq_Choosing_k" format="default"/> respectively for its recommended and <xref target="dualq_Choosing_k" format="default"/>, respectively, fo r its recommended
setting and guidance on alternatives).</t> setting and guidance on alternatives).</t>
<t>There is also a cUrviness parameter, U, which is a small positive <t>There is also a cUrviness parameter, U, which is a small positive
integer. It is likely to take the same hard-coded value for all integer. It is likely to take the same hard-coded value for all
implementations, once experiments have determined a good value. Only implementations, once experiments have determined a good value. Only
U=1 has been used in experiments so far, but results might be even U = 1 has been used in experiments so far, but results might be even
better with U=2 or higher.</t> better with U = 2 or higher.</t>
<t>Notes:</t> <t>Notes:</t>
<ol spacing="normal" type="1"><li anchor="dualq_note_dequeue">The altern <ol spacing="normal" type="1">
ative of applying the <li anchor="dualq_note_dequeue">The alternative of applying the
AQMs at enqueue would shift some processing from the critical time AQMs at enqueue would shift some processing from the critical time
when each packet is dequeued. However, it would also add a whole when each packet is dequeued. However, it would also add a whole
queue of delay to the control signals, making the control loop queue of delay to the control signals, making the control loop
sloppier (for a typical RTT it would double the Classic queue's sloppier (for a typical RTT, it would double the Classic queue's
feedback delay). On a platform where packet timestamping is feedback delay). On a platform where packet timestamping is
feasible, e.g. Linux, it is also easiest to apply the AQMs at feasible, e.g., Linux, it is also easiest to apply the AQMs at
dequeue because that is where queuing time is also measured.</li> dequeue, because that is where queuing time is also measured.</li>
<li anchor="dualq_note_conditional_priority">WRR better isolates <li anchor="dualq_note_conditional_priority">WRR better isolates
the L4S queue from large delay bursts in the Classic queue, but it the L4S queue from large delay bursts in the Classic queue, but it
is slightly less simple than TS-FIFO. If WRR were used, a low is slightly less simple than TS-FIFO. If WRR were used, a low
default Classic weight (e.g. 1/16) would need to be default Classic weight (e.g., 1/16) would need to be
configured in place of the time shift in line 5 of the configured in place of the time-shift in line 5 of the
initialization function (<xref target="dualq_fig_Algo_cred_core_head er" format="default"/>).</li> initialization function (<xref target="dualq_fig_Algo_cred_core_head er" format="default"/>).</li>
<li anchor="dualq_note_step">A step function is shown for <li anchor="dualq_note_step">A step function is shown for
simplicity. A ramp function (see <xref target="dualq_fig_Algo_laqm_c ore" format="default"/> and the discussion around it simplicity. A ramp function (see <xref target="dualq_fig_Algo_laqm_c ore" format="default"/> and the discussion around it
in <xref target="dualq_Ex_algo_pi2-1" format="default"/>) is recomme nded, because in <xref target="dualq_Ex_algo_pi2-1" format="default"/>) is recomme nded, because
it is more general than a step and has the potential to enable L4S it is more general than a step and has the potential to enable L4S
congestion controls to converge more rapidly.</li> congestion controls to converge more rapidly.</li>
<li anchor="dualq_note_non-EWMA">An EWMA is only one possible way <li anchor="dualq_note_non-EWMA">An EWMA is only one possible way
to filter bursts; other more adaptive smoothing methods could be to filter bursts; other more adaptive smoothing methods could be
valid and it might be appropriate to decrease the EWMA faster than valid, and it might be appropriate to decrease the EWMA faster than
it increases, e.g. by using the minimum of the smoothed and it increases, e.g., by using the minimum of the smoothed and
instantaneous queue delays, min(Q_C, qc.time()).</li> instantaneous queue delays, min(Q_C, qc.time()).</li>
</ol> </ol>
</section> </section>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Efficient Implementation of Curvy RED</name> <name>Efficient Implementation of Curvy RED</name>
<t>Although code optimization depends on the platform, the following <t>Although code optimization depends on the platform, the following
notes explain where the design of Curvy RED was particularly motivated notes explain where the design of Curvy RED was particularly motivated
by efficient implementation.</t> by efficient implementation.</t>
<t>The Classic AQM at line 10b calls maxrand(2*U), which gives twice <t>The Classic AQM at line 10b in <xref target="dualq_fig_Algo_Real" for mat="default"/> calls maxrand(2*U), which gives twice
as much curviness as the call to maxrand(U) in the marking function at as much curviness as the call to maxrand(U) in the marking function at
line 5c. This is the trick that implements the square rule in equation line 5c. This is the trick that implements the square rule in equation
(1) (<xref target="dualq_coupled" format="default"/>). This is based on the fact that, (1) (<xref target="dualq_coupled" format="default"/>). This is based on the fact that,
given a number X from 1 to 6, the probability that two dice throws given a number X from 1 to 6, the probability that two dice throws
will both be less than X is the square of the probability that one will both be less than X is the square of the probability that one
throw will be less than X. So, when U=1, the L4S marking function is throw will be less than X.
linear and the Classic dropping function is squared. If U=2, L4S would So, when U = 1, the L4S marking function is
linear and the Classic dropping function is squared. If U = 2, L4S would
be a square function and Classic would be quartic. And so on.</t> be a square function and Classic would be quartic. And so on.</t>
<t>The maxrand(u) function in lines 16-21 simply generates u random <t>The maxrand(u) function in lines 22-27 simply generates u random
numbers and returns the maximum. Typically, maxrand(u) could be run in numbers and returns the maximum. Typically, maxrand(u) could be run in
parallel out of band. For instance, if U=1, the Classic queue would parallel out of band. For instance, if U = 1, the Classic queue would
require the maximum of two random numbers. So, instead of calling require the maximum of two random numbers. So, instead of calling
maxrand(2*U) in-band, the maximum of every pair of values from a maxrand(2*U) in-band, the maximum of every pair of values from a
pseudorandom number generator could be generated out-of-band, and held pseudorandom number generator could be generated out of band and held
in a buffer ready for the Classic queue to consume.</t> in a buffer ready for the Classic queue to consume.</t>
<figure anchor="dualq_fig_Algo_Int"> <figure anchor="dualq_fig_Algo_Int">
<name>Optimised Example Dequeue Pseudocode for DualQ Coupled AQM using Integer Arithmetic</name> <name>Optimised Example Dequeue Pseudocode for DualQ Coupled AQM using Integer Arithmetic</name>
<artwork name="" type="" align="left" alt=""><![CDATA[1: cred_dequeue <sourcecode><![CDATA[
(lq, cq, pkt) { % Couples L4S & Classic queues 1: cred_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: while ( lq.byt() + cq.byt() > 0 ) { 2: while ( lq.byt() + cq.byt() > 0 ) {
3: if ( scheduler() == lq ) { 3: if ( scheduler() == lq ) {
4: lq.dequeue(pkt) % L4S scheduled 4: lq.dequeue(pkt) % L4S scheduled
5: if ((lq.time() > T) OR (Q_C >> (S_L-2) > maxrand(U))) 5: if ((lq.time() > T) OR (Q_C >> (S_L-2) > maxrand(U)))
6: mark(pkt) 6: mark(pkt)
7: } else { 7: } else {
8: cq.dequeue(pkt) % Classic scheduled 8: cq.dequeue(pkt) % Classic scheduled
9: Q_C += (qc.ns() - Q_C) >> g_C % Classic Q EWMA 9: Q_C += (qc.ns() - Q_C) >> g_C % Classic Q EWMA
10: if ( (Q_C >> (S_C-2) ) > maxrand(2*U) ) { 10: if ( (Q_C >> (S_C-2) ) > maxrand(2*U) ) {
11: if ( (ecn(pkt) == 0) { % ECN field = not-ECT 11: if ( (ecn(pkt) == 0) { % ECN field = not-ECT
12: drop(pkt) % Squared drop, redo loop 12: drop(pkt) % Squared drop, redo loop
13: continue % continue to the top of the while loop 13: continue % continue to the top of the while loop
14: } 14: }
15: mark(pkt) 15: mark(pkt)
16: } 16: }
17: } 17: }
18: return(pkt) % return the packet and stop here 18: return(pkt) % return the packet and stop here
19: } 19: }
20: return(NULL) % no packet to dequeue 20: return(NULL) % no packet to dequeue
21: } 21: }
]]></artwork> ]]></sourcecode>
</figure> </figure>
<t>The two ranges, range_L and range_C are expressed as powers of 2 so <t>The two ranges, range_L and range_C, are expressed as powers of 2 so
that division can be implemented as a right bit-shift (&gt;&gt;) in that division can be implemented as a right bit-shift (&gt;&gt;) in
lines 5 and 10 of the integer variant of the pseudocode (<xref target="d ualq_fig_Algo_Int" format="default"/>).</t> lines 5 and 10 of the integer variant of the pseudocode (<xref target="d ualq_fig_Algo_Int" format="default"/>).</t>
<t>For the integer variant of the pseudocode, an integer version of <t>For the integer variant of the pseudocode, an integer version of
the rand() function used at line 25 of the maxrand(function) in <xref ta rget="dualq_fig_Algo_Real" format="default"/> would be arranged to return an int eger the rand() function used at line 25 of the maxrand() function in <xref t arget="dualq_fig_Algo_Real" format="default"/> would be arranged to return an in teger
in the range 0 &lt;= maxrand() &lt; 2^32 (not shown). This would scale in the range 0 &lt;= maxrand() &lt; 2^32 (not shown). This would scale
up all the floating point probabilities in the range [0,1] by up all the floating point probabilities in the range [0,1] by
2^32.</t> 2^32.</t>
<t>Queuing delays are also scaled up by 2^32, but in two stages: i) In <t>Queuing delays are also scaled up by 2^32, but in two stages: i) in
line 9 queuing time qc.ns() is returned in integer nanoseconds, making line 9, queuing time qc.ns() is returned in integer nanoseconds, making
the value about 2^30 times larger than when the units were seconds, the value about 2^30 times larger than when the units were seconds, and
ii) then in lines 5 and 10 an adjustment of -2 to the right bit-shift then
ii) in lines 5 and 10, an adjustment of -2 to the right bit-shift
multiplies the result by 2^2, to complete the scaling by 2^32.</t> multiplies the result by 2^2, to complete the scaling by 2^32.</t>
<t>In line 8 of the initialization function, the EWMA constant gamma <t>In line 8 of the initialization function, the EWMA constant gamma
is represented as an integer power of 2, g_C, so that in line 9 of the is represented as an integer power of 2, g_C, so that in line 9 of the
integer code the division needed to weight the moving average can be integer code (<xref target="dualq_fig_Algo_Int" format="default"/>), the division needed to weight the moving average can be
implemented by a right bit-shift (&gt;&gt; g_C).</t> implemented by a right bit-shift (&gt;&gt; g_C).</t>
</section> </section>
</section> </section>
<section numbered="true" toc="default"> <section numbered="true" toc="default">
<name>Choice of Coupling Factor, k</name> <name>Choice of Coupling Factor, k</name>
<t/> <t/>
<section anchor="dualq_rtt-dependence" numbered="true" toc="default"> <section anchor="dualq_rtt-dependence" numbered="true" toc="default">
<name>RTT-Dependence</name> <name>RTT-Dependence</name>
<t>Where Classic flows compete for the same capacity, their relative <t>Where Classic flows compete for the same capacity, their relative
flow rates depend not only on the congestion probability, but also on flow rates depend not only on the congestion probability but also on
their end-to-end RTT (= base RTT + queue delay). The rates of their end-to-end RTT (= base RTT + queue delay). The rates of
Reno <xref target="RFC5681" format="default"/> flows competing over an A Reno <xref target="RFC5681" format="default"/> flows competing over an A
QM are QM are
roughly inversely proportional to their RTTs. Cubic exhibits similar roughly inversely proportional to their RTTs. CUBIC exhibits similar
RTT-dependence when in Reno-compatibility mode, but it is less RTT-dependence when in Reno-friendly mode, but it is less
RTT-dependent otherwise.</t> RTT-dependent otherwise.</t>
<t>Until the early experiments with the DualQ Coupled AQM, the <t>Until the early experiments with the DualQ Coupled AQM, the
importance of the reasonably large Classic queue in mitigating importance of the reasonably large Classic queue in mitigating
RTT-dependence when the base RTT is low had not been appreciated. RTT-dependence when the base RTT is low had not been appreciated.
Appendix A.1.6 of the L4S ECN protocol <xref target="I-D.ietf-tsvwg-ecn- Appendix <xref target="RFC9331" sectionFormat="bare" section="A.1.6"/>
l4s-id" format="default"/> uses numerical examples to of the L4S ECN Protocol <xref target="RFC9331" format="default"/> uses n
umerical examples to
explain why bloated buffers had concealed the RTT-dependence of explain why bloated buffers had concealed the RTT-dependence of
Classic congestion controls before that time. Then it explains why, Classic congestion controls before that time.
Then, it explains why,
the more that queuing delays have reduced, the more that the more that queuing delays have reduced, the more that
RTT-dependence has surfaced as a potential starvation problem for long RTT-dependence has surfaced as a potential starvation problem for long
RTT flows, when competing against very short RTT flows.</t> RTT flows, when competing against very short RTT flows.</t>
<t>Given that congestion control on end-systems is voluntary, there is <t>Given that congestion control on end systems is voluntary, there is
no reason why it has to be voluntarily RTT-dependent. The no reason why it has to be voluntarily RTT-dependent. The
RTT-dependence of existing Classic traffic cannot be 'undeployed'. RTT-dependence of existing Classic traffic cannot be 'undeployed'.
Therefore, <xref target="I-D.ietf-tsvwg-ecn-l4s-id" format="default"/> r equires L4S Therefore, <xref target="RFC9331" format="default"/> requires L4S
congestion controls to be significantly less RTT-dependent than the congestion controls to be significantly less RTT-dependent than the
standard Reno congestion control <xref target="RFC5681" format="default" />, at standard Reno congestion control <xref target="RFC5681" format="default" />, at
least at low RTT. Then RTT-dependence ought to be no worse than it is least at low RTT. Then RTT-dependence ought to be no worse than it is
with appropriately sized Classic buffers. Following this approach with appropriately sized Classic buffers. Following this approach
means there is no need for network devices to address RTT-dependence, means there is no need for network devices to address RTT-dependence,
although there would be no harm if they did, which per-flow queuing although there would be no harm if they did, which per-flow queuing
inherently does.</t> inherently does.</t>
</section> </section>
<section anchor="dualq_Choosing_k" numbered="true" toc="default"> <section anchor="dualq_Choosing_k" numbered="true" toc="default">
<name>Guidance on Controlling Throughput Equivalence</name> <name>Guidance on Controlling Throughput Equivalence</name>
<t>The coupling factor, k, determines the balance between L4S and <t>The coupling factor, k, determines the balance between L4S and
Classic flow rates (see <xref target="dualq_config" format="default"/> a nd equation Classic flow rates (see <xref target="dualq_config" format="default"/> a nd equation
(1)).</t> (1) in <xref target="dualq_coupled" format="default"/>).</t>
<t>For the public Internet, a coupling factor of k=2 is recommended, <t>For the public Internet, a coupling factor of k = 2 is recommended
and justified below. For scenarios other than the public Internet, a and justified below. For scenarios other than the public Internet, a
good coupling factor can be derived by plugging the appropriate good coupling factor can be derived by plugging the appropriate
numbers into the same working.</t> numbers into the same working.</t>
<t>To summarize the maths below, from equation (7) it can be seen that <t>To summarize the maths below, from equation (7) it can be seen that
choosing k=1.64 would theoretically make L4S throughput roughly the choosing k = 1.64 would theoretically make L4S throughput roughly the
same as Classic, <em>if their actual end-to-end RTTs were the same</em>. same as Classic, <em>if their actual end-to-end RTTs were the same</em>.
However, even if the base RTTs are the same, the actual RTTs are However, even if the base RTTs are the same, the actual RTTs are
unlikely to be the same, because Classic traffic needs a fairly large unlikely to be the same, because Classic traffic needs a fairly large
queue to avoid under-utilization and excess drop. Whereas L4S does queue to avoid underutilization and excess drop, whereas L4S does
not.</t> not.</t>
<t>Therefore, to determine the appropriate coupling factor policy, the <t>Therefore, to determine the appropriate coupling factor policy, the
operator needs to decide at what base RTT it wants L4S and Classic operator needs to decide at what base RTT it wants L4S and Classic
flows to have roughly equal throughput, once the effect of the flows to have roughly equal throughput, once the effect of the
additional Classic queue on Classic throughput has been taken into additional Classic queue on Classic throughput has been taken into
account. With this approach, a network operator can determine a good account. With this approach, a network operator can determine a good
coupling factor without knowing the precise L4S algorithm for reducing coupling factor without knowing the precise L4S algorithm for reducing
RTT-dependence - or even in the absence of any algorithm.</t> RTT-dependence -- or even in the absence of any algorithm.</t>
<t>The following additional terminology will be used, with appropriate <t>The following additional terminology will be used, with appropriate
subscripts:</t> subscripts:</t>
<dl newline="false" spacing="normal"> <dl newline="false" spacing="normal">
<dt>r:</dt> <dt>r:</dt>
<dd>Packet rate [pkt/s]</dd> <dd>Packet rate [pkt/s]</dd>
<dt>R:</dt> <dt>R:</dt>
<dd>RTT [s/round]</dd> <dd>RTT [s/round]</dd>
<dt>p:</dt> <dt>p:</dt>
<dd>ECN marking probability []</dd> <dd>ECN-marking probability []</dd>
</dl> </dl>
<t>On the Classic side, we consider Reno as the most sensitive and <t>On the Classic side, we consider Reno as the most sensitive and
therefore worst-case Classic congestion control. We will also consider therefore worst-case Classic congestion control. We will also consider
Cubic in its Reno-friendly mode ('CReno'), as the most prevalent CUBIC in its Reno-friendly mode ('CReno') as the most prevalent
congestion control, according to the references and analysis in <xref ta rget="PI2param" format="default"/>. In either case, the Classic packet rate in s teady congestion control, according to the references and analysis in <xref ta rget="PI2param" format="default"/>. In either case, the Classic packet rate in s teady
state is given by the well-known square root formula for Reno state is given by the well-known square root formula for Reno
congestion control:</t> congestion control:</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ r_C = 1.22 / ( <sourcecode><![CDATA[
R_C * p_C^0.5) (5)]]></artwork> r_C = 1.22 / (R_C * p_C^0.5) (5)]]></sourcecode>
<t>On the L4S side, we consider the Prague congestion <t>On the L4S side, we consider the Prague congestion
control <xref target="I-D.briscoe-iccrg-prague-congestion-control" forma t="default"/> as the control <xref target="I-D.briscoe-iccrg-prague-congestion-control" forma t="default"/> as the
reference for steady-state dependence on congestion. Prague conforms reference for steady-state dependence on congestion. Prague conforms
to the same equation as DCTCP, but we do not use the equation derived to the same equation as DCTCP, but we do not use the equation derived
in the DCTCP paper, which is only appropriate for step marking. The in the DCTCP paper, which is only appropriate for step marking. The
coupled marking, p_CL, is the appropriate one when considering coupled marking, p_CL, is the appropriate one when considering
throughput equivalence with Classic flows. Unlike step marking, throughput equivalence with Classic flows. Unlike step marking,
coupled markings are inherently spaced out, so we use the formula for coupled markings are inherently spaced out, so we use the formula for
DCTCP packet rate with probabilistic marking derived in Appendix A of DCTCP packet rate with probabilistic marking derived in Appendix A of
<xref target="PI2" format="default"/>. We use the equation without RTT-i ndependence <xref target="PI2" format="default"/>. We use the equation without RTT-i ndependence
enabled, which will be explained later.</t> enabled, which will be explained later.</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ r_L = 2 / (R_L <sourcecode><![CDATA[
* p_CL) (6)]]></artwork> r_L = 2 / (R_L * p_CL) (6)]]></sourcecode>
<t>For packet rate equivalence, we equate the two packet rates and <t>For packet rate equivalence, we equate the two packet rates and
rearrange into the same form as Equation (1), so the two can be rearrange the equation into the same form as equation (1) (copied from < xref target="dualq_coupled" format="default"/>) so the two can be
equated and simplified to produce a formula for a theoretical coupling equated and simplified to produce a formula for a theoretical coupling
factor, which we shall call k*:</t> factor, which we shall call k*:</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ r_c = r_L <sourcecode><![CDATA[
=> p_C = (p_CL/1.64 * R_L/R_C)^2 r_c = r_L
=> p_C = (p_CL/1.64 * R_L/R_C)^2.
p_C = ( p_CL / k )^2 (1) p_C = ( p_CL / k )^2. (1)
k* = 1.64 * (R_C / R_L) (7) k* = 1.64 * (R_C / R_L). (7)
]]></artwork> ]]></sourcecode>
<t>We say that this coupling factor is theoretical, because it is in <t>We say that this coupling factor is theoretical, because it is in
terms of two RTTs, which raises two practical questions: i) for terms of two RTTs, which raises two practical questions: i) for
multiple flows with different RTTs, the RTT for each traffic class multiple flows with different RTTs, the RTT for each traffic class
would have to be derived from the RTTs of all the flows in that class would have to be derived from the RTTs of all the flows in that class
(actually the harmonic mean would be needed); ii) a network node (actually the harmonic mean would be needed) and ii) a network node
cannot easily know the RTT of the flows anyway.</t> cannot easily know the RTT of the flows anyway.</t>
<t>RTT-dependence is caused by window-based congestion control, so it <t>RTT-dependence is caused by window-based congestion control, so it
ought to be reversed there, not in the network. Therefore, we use a ought to be reversed there, not in the network. Therefore, we use a
fixed coupling factor in the network, and reduce RTT-dependence in L4S fixed coupling factor in the network and reduce RTT-dependence in L4S
senders. We cannot expect Classic senders to all be updated to reduce senders. We cannot expect Classic senders to all be updated to reduce
their RTT-dependence. But solely addressing the problem in L4S senders their RTT-dependence. But solely addressing the problem in L4S senders
at least makes RTT-dependence no worse - not just between L4S senders, at least makes RTT-dependence no worse -- not just between L4S senders,
but also between L4S and Classic senders.</t> but also between L4S and Classic senders.</t>
<t>Traditionally, throughput equivalence has been defined for flows <t>Throughput equivalence is defined for flows
under comparable conditions, including with the same base under comparable conditions, including with the same base
RTT <xref target="RFC2914" format="default"/>. So if we assume the same base RTT, RTT <xref target="RFC2914" format="default"/>. So if we assume the same base RTT,
R_b, for comparable flows, we can put both R_C and R_L in terms of R_b, for comparable flows, we can put both R_C and R_L in terms of
R_b.</t> R_b.</t>
<t>We can approximate the L4S RTT to be hardly greater than the base <t>We can approximate the L4S RTT to be hardly greater than the base
RTT, i.e. R_L ~= R_b. And we can replace R_C with (R_b + q_C), RTT, i.e., R_L ~= R_b. And we can replace R_C with (R_b + q_C),
where the Classic queue, q_C, depends on the target queue delay that where the Classic queue, q_C, depends on the target queue delay that
the operator has configured for the Classic AQM.</t> the operator has configured for the Classic AQM.</t>
<t>Taking PI2 as an example Classic AQM, it seems that we could just <t>Taking PI2 as an example Classic AQM, it seems that we could just
take R_C = R_b + target (recommended 15 ms by default in <xref target="d ualq_Ex_algo_pi2-1" format="default"/>). However, target is roughly the queue take R_C = R_b + target (recommended 15 ms by default in <xref target="d ualq_Ex_algo_pi2-1" format="default"/>). However, target is roughly the queue
depth reached by the tips of the sawteeth of a congestion control, not depth reached by the tips of the sawteeth of a congestion control, not
the average <xref target="PI2param" format="default"/>. That is R_max = R_b + the average <xref target="PI2param" format="default"/>. That is R_max = R_b +
target.</t> target.</t>
<t>The position of the average in relation to the max depends on the <t>The position of the average in relation to the max depends on the
amplitude and geometry of the sawteeth. We consider two examples: amplitude and geometry of the sawteeth. We consider two examples:
Reno <xref target="RFC5681" format="default"/>, as the most sensitive wo Reno <xref target="RFC5681" format="default"/>, as the most sensitive wo
rst-case, rst case,
and Cubic <xref target="RFC8312" format="default"/> in its Reno-friendly and CUBIC <xref target="RFC8312" format="default"/> in its Reno-friendly
mode mode
('CReno') as the most prevalent congestion control algorithm on the ('CReno') as the most prevalent congestion control algorithm on the
Internet according to the references in <xref target="PI2param" format=" default"/>. Internet according to the references in <xref target="PI2param" format=" default"/>.
Both are AIMD, so we will generalize using b as the multiplicative Both are Additive Increase Multiplicative Decrease (AIMD), so we will ge
decrease factor (b_r = 0.5 for Reno, b_c = 0.7 for CReno). Then:</t> neralize using b as the multiplicative
<artwork name="" type="" align="left" alt=""><![CDATA[ R_C = (R_max + decrease factor (b_r = 0.5 for Reno, b_c = 0.7 for CReno). Then</t>
b*R_max) / 2 <sourcecode><![CDATA[
= R_max * (1+b)/2 R_C = (R_max + b*R_max) / 2
= R_max * (1+b)/2.
R_reno = 0.75 * (R_b + target); R_creno = 0.85 * (R_b + target). R_reno = 0.75 * (R_b + target); R_creno = 0.85 * (R_b + target).
(8) (8)
]]></artwork> ]]></sourcecode>
<t>Plugging all this into equation (7) we get a fixed coupling factor
<t>Plugging all this into equation (7), at any particular base RTT, R_b,
we get a fixed coupling factor
for each:</t> for each:</t>
<artwork name="" type="" align="left" alt=""><![CDATA[k_reno = 1.64*0.75 <sourcecode><![CDATA[
*(R_b+target)/R_b k_reno = 1.64*0.75*(R_b+target)/R_b
= 1.23*(1 + target/R_b); k_creno = 1.39 * (1 + target/R_b) = 1.23*(1 + target/R_b); k_creno = 1.39 * (1 + target/R_b).
]]></artwork> ]]></sourcecode>
<t>An operator can then choose the base RTT at which it wants <t>An operator can then choose the base RTT at which it wants
throughput to be equivalent. For instance, if we recommend that the throughput to be equivalent. For instance, if we recommend that the
operator chooses R_b = 25 ms, as a typical base RTT between Internet operator chooses R_b = 25 ms, as a typical base RTT between Internet
users and CDNs <xref target="PI2param" format="default"/>, then these co upling users and CDNs <xref target="PI2param" format="default"/>, then these co upling
factors become:</t> factors become:</t>
<artwork name="" type="" align="left" alt=""><![CDATA[k_reno = 1.23 * (1 <sourcecode><![CDATA[
+ 15/25) k_creno = 1.39 * (1 + 15/25) k_reno = 1.23 * (1 + 15/25) k_creno = 1.39 * (1 + 15/25)
= 1.97 = 2.22 = 1.97 = 2.22
~= 2 ~= 2 (9) ~= 2. ~= 2. (9)
]]></artwork> ]]></sourcecode>
<t>The approximation is relevant to any of the above example DualQ <t>The approximation is relevant to any of the above example DualQ
Coupled algorithms, which use a coupling factor that is an integer Coupled algorithms, which use a coupling factor that is an integer
power of 2 to aid efficient implementation. It also fits best to the power of 2 to aid efficient implementation. It also fits best for the
worst case (Reno).</t> worst case (Reno).</t>
<t>To check the outcome of this coupling factor, we can express the <t>To check the outcome of this coupling factor, we can express the
ratio of L4S to Classic throughput by substituting from their rate ratio of L4S to Classic throughput by substituting from their rate
equations (5) and (6), then also substituting for p_C in terms of equations (5) and (6), then also substituting for p_C in terms of
p_CL, using equation (1) with k=2 as just determined for the p_CL using equation (1) with k = 2 as just determined for the
Internet:</t> Internet:</t>
<artwork name="" type="" align="left" alt=""><![CDATA[r_L / r_C = 2 (R_ <sourcecode><![CDATA[
C * p_C^0.5) / 1.22 (R_L * p_CL) r_L / r_C = 2 (R_C * p_C^0.5) / 1.22 (R_L * p_CL)
= (R_C * p_CL) / (1.22 * R_L * p_CL) = (R_C * p_CL) / (1.22 * R_L * p_CL)
= R_C / (1.22 * R_L) (10) = R_C / (1.22 * R_L). (10)
]]></artwork> ]]></sourcecode>
<t>As an example, we can then consider single competing CReno and <t>As an example, we can then consider single competing CReno and
Prague flows, by expressing both their RTTs in (10) in terms of their Prague flows, by expressing both their RTTs in (10) in terms of their
base RTTs, R_bC and R_bL. So R_C is replaced by equation (8) for base RTTs, R_bC and R_bL. So R_C is replaced by equation (8) for
CReno. And R_L is replaced by the max() function below, which CReno. And R_L is replaced by the max() function below, which
represents the effective RTT of the current Prague congestion represents the effective RTT of the current Prague congestion
control <xref target="I-D.briscoe-iccrg-prague-congestion-control" forma t="default"/> in its control <xref target="I-D.briscoe-iccrg-prague-congestion-control" forma t="default"/> in its
(default) RTT-independent mode, because it sets a floor to the (default) RTT-independent mode, because it sets a floor to the
effective RTT that it uses for additive increase:</t> effective RTT that it uses for additive increase:</t>
<artwork name="" type="" align="left" alt=""><![CDATA[ ~= 0.85 <sourcecode><![CDATA[
* (R_bC + target) / (1.22 * max(R_bL, R_typ)) r_L / r_C ~= 0.85 * (R_bC + target) / (1.22 * max(R_bL, R_typ))
~= (R_bC + target) / (1.4 * max(R_bL, R_typ)) ~= (R_bC + target) / (1.4 * max(R_bL, R_typ)).
]]></artwork> ]]></sourcecode>
<t>It can be seen that, for base RTTs below target (15 ms), both the <t>It can be seen that, for base RTTs below target (15 ms), both the
numerator and the denominator plateau, which has the desired effect of numerator and the denominator plateau, which has the desired effect of
limiting RTT-dependence.</t> limiting RTT-dependence.</t>
<t>At the start of the above derivations, an explanation was promised <t>At the start of the above derivations, an explanation was promised
for why the L4S throughput equation in equation (6) did not need to for why the L4S throughput equation in equation (6) did not need to
model RTT-independence. This is because we only use one point - at the model RTT-independence. This is because we only use one point -- at the
typical base RTT where the operator chooses to calculate the coupling typical base RTT where the operator chooses to calculate the coupling
factor. Then, throughput equivalence will at least hold at that chosen factor. Then throughput equivalence will at least hold at that chosen
point. Nonetheless, assuming Prague senders implement RTT-independence point. Nonetheless, assuming Prague senders implement RTT-independence
over a range of RTTs below this, the throughput equivalence will then over a range of RTTs below this, the throughput equivalence will then
extend over that range as well.</t> extend over that range as well.</t>
<t>Congestion control designers can choose different ways to reduce <t>Congestion control designers can choose different ways to reduce
RTT-dependence. And each operator can make a policy choice to decide RTT-dependence. And each operator can make a policy choice to decide
on a different base RTT, and therefore a different k, at which it on a different base RTT, and therefore a different k, at which it
wants throughput equivalence. Nonetheless, for the Internet, it makes wants throughput equivalence. Nonetheless, for the Internet, it makes
sense to choose what is believed to be the typical RTT most users sense to choose what is believed to be the typical RTT most users
experience, because a Classic AQM's target queuing delay is also experience, because a Classic AQM's target queuing delay is also
derived from a typical RTT for the Internet.</t> derived from a typical RTT for the Internet.</t>
<t>As a non-Internet example, for localized traffic from a particular <t>As a non-Internet example, for localized traffic from a particular
ISP's data centre, using the measured RTTs, it was calculated that a ISP's data centre, using the measured RTTs, it was calculated that a
value of k = 8 would achieve throughput equivalence, and experiments value of k = 8 would achieve throughput equivalence, and experiments
verified the formula very closely.</t> verified the formula very closely.</t>
<t>But, for a typical mix of RTTs across the general Internet, a value <t>But, for a typical mix of RTTs across the general Internet, a value
of k=2 is recommended as a good workable compromise.</t> of k = 2 is recommended as a good workable compromise.</t>
</section> </section>
</section> </section>
<!-- <section title="Open Issues">
<t>Minor open issues are tagged '{ToDo}' at the appropriate point in the
document. Major open issues are listed below:<list>
<t>None</t>
</list></t>
</section>
<section title="Change Log (to be Deleted before Publication)">
<t>A detailed version history can be accessed at
&lt;http://datatracker.ietf.org/doc/draft-briscoe-aqm-ecn-roadmap/history/
&gt;</t>
<t><list style="hanging">
<t hangText="From briscoe-...-00 to briscoe-...-01:">Technical
changes:<list style="symbols">
<t/>
</list>Editorial changes:<list style="symbols">
<t/>
</list></t>
</list></t>
</section>
<section numbered="false" toc="default"> <section numbered="false" toc="default">
<name>Acknowledgements</name> <name>Acknowledgements</name>
<t>Thanks to Anil Agarwal, Sowmini Varadhan, Gabi Bracha, Nicolas Kuhn, <t>Thanks to <contact fullname="Anil Agarwal"/>, <contact
Greg Skinner, Tom Henderson, David Pullen, Mirja Kuehlewind, Gorry fullname="Sowmini Varadhan"/>, <contact fullname="Gabi Bracha"/>,
Fairhurst, Pete Heist, Ermin Sakic and Martin Duke for detailed review <contact fullname="Nicolas Kuhn"/>, <contact fullname="Greg Skinner"/>,
comments particularly of the appendices and suggestions on how to make <contact fullname="Tom Henderson"/>, <contact fullname="David Pullen"/>,
the explanations clearer. Thanks also to Tom Henderson for insights on <contact fullname="Mirja Kühlewind"/>, <contact fullname="Gorry
the choice of schedulers and queue delay measurement techniques. And Fairhurst"/>, <contact fullname="Pete Heist"/>, <contact fullname="Ermin
thanks to the area reviewers Christer Holmberg, Lars Eggert and Roman Sakic"/>, and <contact fullname="Martin Duke"/> for detailed review
Danyliw.</t> comments, particularly of the appendices, and suggestions on how to make
<t>The early contributions of Koen De Schepper, Bob Briscoe, Olga the explanations clearer. Thanks also to <contact fullname="Tom
Bondarenko and Inton Tsang were part-funded by the European Community Henderson"/> for insight on the choice of schedulers and queue delay
measurement techniques. And thanks to the area reviewers <contact
fullname="Christer Holmberg"/>, <contact fullname="Lars Eggert"/>, and
<contact fullname="Roman Danyliw"/>.</t>
<t>The early contributions of <contact fullname="Koen De Schepper"/>, <con
tact fullname="Bob Briscoe"/>, <contact fullname="Olga
Bondarenko"/>, and <contact fullname="Inton Tsang"/> were partly funded by
the European Community
under its Seventh Framework Programme through the Reducing Internet under its Seventh Framework Programme through the Reducing Internet
Transport Latency (RITE) project (ICT-317700). Contributions of Koen De Transport Latency (RITE) project (ICT-317700). Contributions of <contact f
Schepper and Olivier Tilmans were also part-funded by the 5Growth and ullname="Koen De
DAEMON EU H2020 projects. Bob Briscoe's contribution was also Schepper"/> and <contact fullname="Olivier Tilmans"/> were also partly fun
part-funded by the Comcast Innovation Fund and the Research Council of ded by the 5Growth and
DAEMON EU H2020 projects. <contact fullname="Bob Briscoe"/>'s contribution
was also
partly funded by the Comcast Innovation Fund and the Research Council of
Norway through the TimeIn project. The views expressed here are solely Norway through the TimeIn project. The views expressed here are solely
those of the authors.</t> those of the authors.</t>
</section> </section>
<section numbered="false" toc="default"> <section numbered="false" toc="default">
<name>Contributors</name> <name>Contributors</name>
<t>The following contributed implementations and evaluations that <t>The following contributed implementations and evaluations that
validated and helped to improve this specification:</t> validated and helped to improve this specification:</t>
<ul empty="true" spacing="normal"> <t><contact fullname="Olga Albisser"/> &lt;olga@albisser.org&gt; of Simu
<li>Olga Albisser &lt;olga@albisser.org&gt; of Simula Research Lab, la Research Lab,
Norway (Olga Bondarenko during early drafts) implemented the Norway (Olga Bondarenko during early draft versions) implemented the
prototype DualPI2 AQM for Linux with Koen De Schepper and conducted prototype DualPI2 AQM for Linux with Koen De Schepper and conducted
extensive evaluations as well as implementing the live performance extensive evaluations as well as implementing the live performance
visualization GUI <xref target="L4Sdemo16" format="default"/>.</li> visualization GUI <xref target="L4Sdemo16" format="default"/>.</t>
<li>Olivier Tilmans &lt;olivier.tilmans@nokia-bell-labs.com&gt; of <t><contact fullname="Olivier Tilmans"/> &lt;olivier.tilmans@nokia-bell-
labs.com&gt; of
Nokia Bell Labs, Belgium prepared and maintains the Linux Nokia Bell Labs, Belgium prepared and maintains the Linux
implementation of DualPI2 for upstreaming.</li> implementation of DualPI2 for upstreaming.</t>
<li>Shravya K.S. wrote a model for the ns-3 simulator based on the <t><contact fullname="Shravya K.S."/> wrote a model for the ns-3 simulat
-01 version of this Internet-Draft. Based on this initial work, Tom or based on draft-ietf-tsvwg-aqm-dualq-coupled-01 (a draft version of this docum
Henderson &lt;tomh@tomh.org&gt; updated that earlier model and ent). Based on this initial work, <contact fullname="Tom
created a model for the DualQ variant specified as part of the Low Henderson"/> &lt;tomh@tomh.org&gt; updated that earlier model and
Latency DOCSIS specification, as well as conducting extensive created a model for the DualQ variant specified as part of the Low Lat
evaluations.</li> ency
<li>Ing Jyh (Inton) Tsang of Nokia, Belgium built the End-to-End Data DOCSIS specification, as well as conducting extensive
evaluations.</t>
<t><contact fullname="Ing Jyh (Inton) Tsang"/> of Nokia, Belgium built t
he End-to-End Data
Centre to the Home broadband testbed on which DualQ Coupled AQM Centre to the Home broadband testbed on which DualQ Coupled AQM
implementations were tested.</li> implementations were tested.</t>
</ul>
</section> </section>
</back> </back>
</rfc> </rfc>
 End of changes. 551 change blocks. 
2127 lines changed or deleted 1308 lines changed or added

This html diff was produced by rfcdiff 1.48.