Review: MACAW: A Media Access Protocol for Wireless LAN's

Paper: MACAW: A Media Access Protocol for Wireless LAN's

Name: Robert Danek.
Course: CS2209, Fall '06.

This paper explores an existing design for a wireless media access
protocol, called MACA (Multiple Access, Collision Avoidance). It
summarizes a number of problems and drawbacks of this protocol, and then
goes on to propose a new protocol that the authors dub MACAW. The
authors explain the details of MACAW in an iterative manner, by first
specifying a particular problem they want to solve, and then showing
what feature they added to MACAW to solve the problem.

Before describing the details of either MACA or MACAW, the authors first
explain why a variant of the standard CSMA protocol used by regular
Ethernets would not work in a wireless environment. The main problems
with using carrier sense technology are the "hidden terminal" problem,
and the "exposed terminal" problem.

The MACA protocol attempts to solve the problems of CSMA by using
Request-To-Send (RTS) and Clear-To-Send (CTS) messages between hosts
that wish to communicate. The way MACA handles collisions, via its
exponential backoff scheme, is the first thing that the authors
criticize. There is a certain amount of unfairness associated with this
scheme. To rectify this, in MACAW, backoff counters are propagated via
the packet header of messages to other hosts. The authors also improve
the efficiency of the scheme by modifying the amount that a host's
backoff counter is incremented/decremented when it's changed.

Another issue that the authors have with MACA is the lack of
acknowledgements that occur at the link layer. In its current form, if
data is lost in MACA, the recovery has to occur at the transport layer.
This can take an order of magnitude longer than if the recovery was done
in the link layer. To solve this, the authors add ACK messages to the
MACAW protocol.

The introduction of ACKs into MACAW resulted in a new problem. Whereas
before it might be safe for a host to send even though someone else was
sending, now there exists a risk that the same host could cause a
collision to occur with an ACK. To solve this, the authors introduce a
new Data-Sending (DS) message that tells other hosts in the vicinity not
to send at the same time.

The authors go on to explore a few more very specific scenarios and the
problems that the scenarios demonstrate. As a result, another message,
Request-for-Request-to-Send (RRTS) is added to the protocol, and further
modifications are made to the collision backoff scheme.

The paper concludes by evaluating the MACAW protocol, and describing
some future design issues. The performance results reported were that at
low loads, MACAW performs worse than MACA, but under conditions of
greater congestion MACAW performs better and is more fair. The design
issues mentioned include problems with multicast, and the possibility of
using token-based schemes.

Though the authors point out that congestion is location dependent in
wireless networks, they never seem to fully justify why building
fairness into the protocol is a necessity. Will wireless networks always
be operating under conditions of high congestion, which would
necessitate fairness? This question never seems to be asked or answered.
The other factor that would necessitate fairness is the presence of
noise on the network, corrupting packets and making it appear as if
there were high congestion. They simulate noise in their experiments,
but they don't provide justification for the noise levels they choose.
Received on Thu Sep 21 2006 - 10:43:36 EDT