Multi-hop Ad Hoc Wireless CommunicationINRIA+33-169-335-511Emmanuel.Baccelli@inria.frhttp://www.emmanuelbaccelli.org/Futurewei+1-408-330-5305charlie.perkins@huawei.com
Internet
Mobile Ad-hoc Networks (MANET)I-DInternet Draft
This document describes some characteristics
of communication between nodes in a multi-hop ad hoc
wireless network. These are not requirements in the
sense usually understood as applying to formulation of
a requirements document.
Nevertheless, protocol engineers and system analysts
involved with designing solutions for ad hoc networks
must maintain awareness of these characteristics.
The goal of this document is to
describe some aspects of multi-hop ad hoc wireless
communication. Experience gathered with
shows that this type of
communication presents specific challenges. This
document briefly describes these challenges, which one
should maintain awareness of, when designing Internet
protocols for ad hoc networks.
For the purposes of this document, a multi-hop ad hoc wireless
network will be considered to be a collection of devices
that each have a radio transceiver, that are using the same
physical and medium access protocols, that are moreover configured
to self-organize and provide store-and-forward functionality
on top of these protocols as needed to enable communications.
The devices providing network connectivity are considered to
be routers. Other non-routing wireless devices, if present
in the ad hoc network, are considered to be "end-hosts".
The considerations in this document apply equally to routers
or end-hosts; we use the term "node" to refer to any such
network device in the ad hoc network.
An example of multi-hop ad hoc wireless network is a wireless
community network such as Funkfeuer
or Freifunk , that consists in routers
running OLSR on 802.11 in ad hoc mode
with the same ESSID at link layer. Multi-hop ad hoc wireless
networks may also run on link layers other than 802.11.
Note however that simple hosts communicating through an access
point with 802.11 in infrastructure mode do not form a multi-hop
ad hoc wireless network, since the central role of the access point
is determined a priori, and since nodes other than the access point
do not generally provide store-and-forward functionality.
Let A and B be two nodes in a multi-hop ad hoc wireless network N.
Suppose that, when node A transmits a packet through its
interface on network N, that packet is correctly received
by node B without requiring storage and/or forwarding by
any other device. We will then say that B "hears" packets from A.
Note that therefore, when B can hear IP packets from A, the TTL of
the IP packet heard by B will be precisely the same as it
was when A transmitted that packet.
Let S be the set of nodes that can hear packets transmitted
by node A through its interface on network N. The following
section gathers common characteristics concerning packet
transmission over such networks, which were observed through
experience with
.
First, there is no guarantee that a node C within S can,
symmetrically, send IP packets directly to node A. In other
words, even though C can "hear" packets from A (since it is
a member of set S), there is no guarantee that A can "hear"
packets from C. Thus, multi-hop ad hoc wireless communications
may be "asymmetric". Such cases are not uncommon.
Second, there is no guarantee that, as a set, S is at all stable, i.e.
the membership of set S may in fact change at any rate, at any time.
Thus, multi-hop ad hoc wireless communications may be "time-variant".
Such variations are not unusual in multi-hop ad hoc wireless networks
due to variability of the wireless medium, and to node mobility.
Now, conversely, let V be the set of nodes from which A can
directly receive packets -- in other words, A can "hear" packets
from any node in set V. Suppose that node A is
communicating at time t0 through its interface on network N.
As a consequence of time variation and asymmetry,
we observe that A:
cannot assume that S = V,
cannot assume that S and/or V are unchanged at time t1 later than t0.
Furthermore, transitivity is not guaranteed over multi-hop ad hoc
wireless networks. Indeed, let's assume that, through their
respective interfaces within network N:
node B and node A can hear each other (i.e. node B is a member of
sets S and V), and,
node A and node C can also hear each other (i.e. node C is a
also a member of sets S and V).
These assumptions do not imply that node B can hear node C, nor that
node C can hear node B (through their interface on network N).
Such "non-transitivity" is not uncommon on multi-hop ad hoc
wireless networks.
In a nutshell: multi-hop ad hoc wireless communications can be
asymmetric, non-transitive, and time-varying.
presents an abstract description of some
common characteristics concerning packet transmission over multi-hop
ad hoc wireless networks. This section describes practical examples,
which illustrate the characteristics listed in
as well as other common effects.
Wireless communication links are subject to limitations to the
distance across which they may be established. The range-limitation
factor creates specific problems on multi-hop ad hoc wireless
networks. In this context, it is not uncommon that the radio ranges
of several nodes partially overlap. Such partial overlap causes
communication to be non-transitive and/or asymmetric, as described
in .
For example, as depicted in Figure 1, it may happen that a node
B hears a node A which transmits at high power, whereas B transmits
at lower power. In such cases, B can hear A, but A cannot hear B.
This examplifies the asymmetry in multi-hop ad hoc wireless
communications as defined in .
Another example, depicted in Figure 2, is known as the "hidden node"
problem. Even though the nodes all have equal power for their radio
transmissions, they cannot all reach one another. In the figure,
nodes A and B can hear each other, and A and C can also hear each
other. On the other hand, nodes B and C cannot hear each other. When
nodes B and C try to communicate with node A at the same time, their
radio signals collide. Node A will only be able to detect noise, and
may even be unable to determine the source of the noise. The hidden
terminal problem illustrates the property of non-transitivity in
multi-hop ad hoc wireless communications as described
in .
Another situation, shown in Figure 3, is known as the
"exposed node" problem. In the figure, node A is transmitting
(to node B). As shown, node C cannot communicate properly with
node D, because of the on-going transmission of node A, polluting
C's radio-range. Node C cannot hear D, but node D can
hear C because D is outside A's radio range.
Node C is then called an "exposed node", because it is exposed to
co-channel interference from node A and thereby prevented from
exchanging protocol messages to enable transmitting data to node D
-- even though the transmission would be successful and would not
interfere with the reception of data sent from node A to node B.
Hidden and exposed node situations are not uncommon
in multi-hop ad hoc wireless networks. Problems with asymmetric
links may also arise for reasons other than power inequality
(e.g., multipath interference).
Such problems are often resolved by specific mechanisms below the
IP layer.
However, depending the link layer technology in use and the
position of the nodes, such problems due to
range-limitation and partial overlap may affect the IP layer.
Besides radio range limitations, wireless communications are
affected by irregularities in the shape of the geographical area
over which nodes may effectively communicate (see for instance
, ).
For example, even omnidirectional wireless transmission is
typically non-isotropic (i.e. non-circular).
Signal strength often suffers frequent and significant variations,
which are not a simple function of distance. Instead, it is a
complex function of the environment
including obstacles, weather conditions, interference, and
other factors that change over time. The analytical
formulation of such variation is often considered intractable.
These irregularities also cause communications on multi-hop ad hoc
wireless networks to be non-transitive, asymmetric, or time-varying,
as described in , and may impact the IP layer.
There may be no indication to IP when a previously established
communication channel becomes unusable; "link down" triggers
are generally absent in multi-hop ad hoc wireless networks.
Many terms have been used in the past to describe the relationship of
nodes in a multi-hop ad hoc wireless network based on their ability
to send or receive packets to/from each other. The terms used in
this document have been selected because the authors believe (or at
least hope) they are unambiguous, with respect to the
goal of this document (see ).
Nevertheless, here are a few other terms that describe the same
relationship between nodes in multi-hop ad hoc wireless networks.
In the following, let network N be, again, a multi-hop ad hoc
wireless network. Let the set S be, as before, the set of
nodes that can directly receive packets transmitted by node A
through its interface on network N. In other words, any node B
belonging to S can "hear" packets transmitted by A. Then,
due to the asymmetry characteristic of wireless links:
- We may say that node B is reachable from node A. In this
terminology, there is no guarantee that node A is reachable from
node B, even if node B is reachable from node A.
- We may say that node A has a link to node B. In this terminology,
there is no guarantee that node B has a link to node A, even if
node A has a link to node B.
- We may say that node B is adjacent to node A. In this terminology,
there is no guarantee that node A is adjacent to node B, even
if node B is adjacent to node A.
- We may say that node B is downstream from node A. In this
terminology, there is no guarantee that node A is downstream from
node B, even if node B is downstream from node A.
- We may say that node B is a neighbor of node A. In this
terminology, there is no guarantee that node A is a neighbor
of node B, even if node B a neighbor of node A. As it happens,
the terminology for "neighborhood" is quite confusing for
asymmetric links. When B can hear signals from A, but A
cannot hear B, it is not clear whether B should be considered
a neighbor of A at all, since A would not necessarily be
aware that B was a neighbor. Perhaps it is best to avoid
the "neighbor" terminology except for symmetric links.
This list of alternative terminologies is given here for illustrative
purposes only, and is not suggested to be complete or even
representative of the breadth of terminologies that have been
used in various ways to explain the properties mentioned in
.
While the above-described characteristics of packet transmission over
multi-hop ad hoc wireless networks are not the default characteristics
expected by IP , it is possible and desirable to run IP
over such networks. Over the last decade, numerous example of such networks
running IP have turned up, including ,
among many other examples. IP has been running over
such networks through the use of:
specific IP interface configuration, as described in
RFC 5889 , or
specific routing protocols running on these interfaces,
such as
for instance.
Thus, even though the physical effects described in this document impose
significant requirements for enhancing the robustness of protocol designs
for routing and topology management, the experience in the projects
described in the cited references shows that useful networks can be
designed and operated using well-understood techniques. It is thus
possible to shield other IP protocols running on other interfaces from
the unusual characteristics experienced over multi-hop ad hoc wireless
networks. Note however that some protocols are still more appropriate
than others when interfaces to multi-hop ad hoc wireless networks are
involved in the communication. For instance, for applications written
to run over both UDP and TCP, the latter choice may be preferred in
situations with relatively high packet loss rates. But such choices
must be based on application requirements.
This document does not have any security considerations.
This document does not have any IANA actions.
A DoD perspective on mobile ad hoc networkshttp://www.funkfeuer.atAustria Wireless Community Network,
http://www.funkfeuer.athttps://map.funkfeuer.at/wien/Mobile Ad hoc Networking: Routing Technology
for Dynamic, Wireless NetworksThe Mistaken Axioms of Wireless-Network ResearchFreifunk Wireless Community Networkshttp://www.freifunk.net
This document stems from discussions with the following people,
in alphabetical order:
Jari Arkko,
Teco Boot,
Carlos Jesus Bernardos Cano,
Ian Chakeres,
Thomas Clausen,
Christopher Dearlove,
Ralph Droms,
Ulrich Herberg,
Paul Lambert,
Kenichi Mase,
Thomas Narten,
Erik Nordmark,
Alexandru Petrescu,
Stan Ratliff,
Zach Shelby,
Shubhranshu Singh,
Fred Templin,
Dave Thaler,
Mark Townsley,
Ronald Velt in't,
and
Seung Yi.