From: Kate Everitt (everitt@cs.washington.edu)
Date: Mon Nov 22 2004 - 07:59:16 PST
This paper discusses the MACAW wireless access protocol, which is based on
MACA. One big challenge in working in this area is that some pads or
senders cannot hear others, even though it will affect their transmission,
and the congestion is location dependant. This means that carrier sense
congestion doesn.t make sense, because the carrier does not know about
all the transmissions that could affect congestion at the receiver. The
author also notes that it is required to share the backoff counter so that
one pad doesn.t increasingly .win. the channel because it has a lower
backoff (as it isn.t experiencing the congestion to the same extent.) This
fixes the problem of MACA, which was that exponential backoff was unfair
and unstable. The author uses RTS-CTS-DATA-ACK to deal with the fact that
each node can.t hear the traffic at all the other nodes, so it has to
infer a transmission given the information it has.
This algorithm gives a 37% throughput improvement in simulation, and it
looks like the overhead is worth it because it improves congestion
control, which is a critical problem in wireless.
However, it doesn.t do a good job of describing how this would apply to
more heterogeneous senders, which is the situation today. It doesn.t deal
at all with the problem of overlapping senders, but it is a bit
unreasonable to expect someone putting in a wireless network to make sure
that this will not occur.
This archive was generated by hypermail 2.1.6 : Mon Nov 22 2004 - 07:59:16 PST