[Cu-wireless] 802.11 & multihop networks

[David Young]

Here is my explanation of RTS/CTS packets in 802.11, and some references
to research concerning 802.11 in multihop networks such as C-U Wireless
is building. The research results are either cause for despair, if you're
into that kind of thing, or cause for some creative problem-solving.

Stations in an 802.11 network may use RTS/CTS to quiet their neighbors
before transmitting or receiving, respectively. The 802.11 committee
intended RTS/CTS to improve performance in networks where every station
cannot hear every other, but there are so-called "hidden nodes". Just
for an example, say we have nodes A, B, and C:

   [A]         [B]          [C]

A and B are in radio range. A and C are not in radio range, however B
and C are. Say that both A and C have packets queued for B. Suppose for
simplicity's sake say that A senses the medium before C, and detects that
it is quiet. A begins its transmission to B. Before A's transmission
finishes, C also senses the medium, detects that it is quiet, and
begins to transmit.  A's and C's packets collide at B, so B does not
send an acknowledgement to either. Both A's and C's packets have to be
re-sent. In this way, a lot of precious bandwidth can be wasted.

RTS and CTS are short messages that help coordinate hidden nodes to
conserve bandwidth.  When A detects that the medium is quiet, it sends B
an RTS. The content of the RTS is "I'm talking to B for X microseconds,
all y'all shut up." If B receives the RTS, B sends back a CTS whose
content is "A is talking to me for X microseconds; all y'all shut
up." This shuts-up B's neighbors, including C, until A sends the Data
packet and B sends an ACK.

Great, huh? Well, sort of. Remember that A's neighbors are shut up
for the duration of A's CTS/Data/ACK transaction with B. In that
time, none of A's neighbors could have transmitted to A, however,
they could have transmitted to other stations in the network
while A sent a Data packet to B. The paper called "Solution ot
the Exposed Node Problem of 802.11 in Wireless Ad-Hoc Networks"
<http://www.cs.iastate.edu/~vel/research/E-MAC.pdf> describes a
backwards-compatible modification to 802.11 that solves the problem
in simulations.

Another deficiency of RTS/CTS is that with its use, network throughput
does not settle to a maximum figure as network load approaches infinity,
but it actually drops to zero!  The paper "RTS/CTS-Induced Congestion
in Ad Hoc Wireless LANs" <http://people.bu.edu/staro/wcnc-ray.pdf>
discusses the problem and describes a backwards-compatible modification
to 802.11 that solves it in simulations.

TCP/IP and 802.11 wireless do not always get along.  In simulations,
when two TCP streams are channeled into the same 802.11 receiver, one
will dominate all of the bandwidth under certain conditions. A paper
on this is called "Does IEEE 802.11 MAC Protocol Work Well in Multihop
Wireless Ad Hoc Networks"; search for the title with Google.

Dave

-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933