Review - MACAW

Review of the paper "MACAW: A Media Access Protocol for Wireless LAN's"
by Ivan Hernández

The paper shows the design and implementation of a new MAC protocol
for wireless LANs: MACAW. This is build based on MACA (Multiple Access
Collision Avoidance) protocol, the enhancements made to this protocol
are highly influenced by four keen-sighted observations about the
wireless LAN behavior and characteristics and by two goals: deliver
high network utilization and provide fair access to the media.

The authors describe the operation of MACA, the RTS-CTS-DATA exchange,
the purpose of each message, and algorithm for the retransmission time
(Binary Exponential Backoff). They highlight that the RTS-CTS exchange
enables nearby stations to avoid collision at the receivers. They also
show how these protocol solves the hidden and exposed terminal
problems.

At this point, the authors describe their modifications to MACA, and
how these, achieve the aimed goals:

a) The first change is the replacement of the backoff algorithm, the
new approach has two characteristics: it is shared by all the nearby
stations and does not has extreme adjustments; in addition, they
implement separate queues for each stream and run the new backoff
algorithm independently. As a result of this modifications, the MAC
protocol increments its fairness access level.

b) The authors identify a weak trait with MACA: if there are collision
the error has to be recovered by the transport layer, which can be
slow due the time tolerance of TCP. So the authors, as an
optimization, add the ACK message. In the simulations, we have a good
trade-off between the overhead introduced by the message and the
gained throughput when noise is present.

c) The authors describe an scenario when a station B is not able to
listen a RST because another nearby stream is taking place, and thus
provokes collisions to any packet directed to B. This will result in
an increment in B backoff counter and in a decrease in its chances to
transmit. The authors' solution is add a message (DS), that will
inform all nearby stations about the duration of a new stream, so each
stations that is able to listen this message will wait until the end of
this transmission in order to try to gain the medium.

d) One interesting improvement to MACA is the capacity of a receiver,
that because of a nearby stream can not answer a RTS, to Request for a
RTS (RRTS message) as soon as the channel is available.

Finally, they evaluate MACAW on a pair of complex configurations, the
results are quite interesting compared with MACA: the throughput has a
significant increment and the division of the throughput among the
streams in any given cell is fair.

I really liked this paper, all the presented ideas and scenarios are
clearly expressed and exemplify. I found interesting its approach to
achieve fair access to the media: the stations must have the "same"
knowledge about the state/congestion of the network, furthermore, this
knowledge sensitive to the station's locality. Some of the problems
stated on the paper (Figs. 6, 7 and the leakage problem) are product
of the single channel approach, nowadays we use multichannel APs that
just get rid of the problems that could cause an adjacent cell. About
the multicast unsolved problem, it looks like a problem by far out of
the reach of this paper, i.e. this problem per se could be a research
area/paper. Finally, even when it was simulated on a extremely poor
model this paper presents huge improvements to the original MAC
protocol.
Received on Mon Sep 25 2006 - 16:38:51 EDT