From: Michael J Cafarella (mjc@cs.washington.edu)
Date: Sun Nov 21 2004 - 23:26:02 PST
MACAW: A Media Access Protocol for Wireless LANs
By Bharghavan, Demers, Shenker, and Zhang
Review by Michael Cafarella
CSE561
November 22, 2004
The authors present the MACAW algorithm, an extension to MACA. Local-area
wireless systems pose a number of unusual networking problems. These problems
usually involve guaranteeing basic network capabilities (such as fairness,
high throughput, and support for varying traffic rates) in the face of unusual
media access properties and signal interference. The MACAW system uses a system
of "send announcements" and "send permission" packets to make everything work.
I liked this paper a lot, because with a very simple framework and a few
straightforward extensions it is able to reach most of the goals at hand. The
big idea of the paper is a system of RTS and CTS packets. Traditional carrier
sense doesn't work with wireless networks, because whatever the transmitter
senses does not necessarily reflect the receiver's media environment. So
instead a transmitter sends a Request-to-Send packet, which is then
perhaps acknowledged by the receiver's CTS packet. This exchange not only
tells the transmitter when it is permissible to send, it also announces to
other nearby transmitters when they must be quiet so the exchange can
take place uninterrupted.
There are a number of extensions, including: every transmitter has a backoff
counter before retrying a RTS packet; these counters are synchronized across
transmitters; the backoff period is increased multiplicatively and
decreased additively; throughput starvation is prevented through use of
RRTS packet, which prompts certain senders for traffic. A nice attribute
of all these extensions is that they are quite natural and build the system's
abilities piece-by-piece. Other systems (for example, the I3 paper) seem
to get all their power from drastic add-ons rather than the basic primitive.
Unfortunately, all the extensions make the algorithm somewhat hard to analyze.
The authors give performance numbers, but no formal framework for discussing
the protocol. The paper itself presents a number of encountered problems,
but it's hard to say whether there could be others.
Further, the usage model is very tightly defined. Cells are very small
and non-overlapping, and cells do not move. The non-overlapping rule,
especially, seems too restrictive for most environments.
This paper is very relevant to the current day, when wireless systems are
used everywhere. It may be true that 802.11b already solves many of the
problems here, so the paper may be more historically interesting rather than
directly useful. I would add to the paper by broadening the range of networks
that MACAW handles, and by making the analysis more rigorous. These two
projects would make the paper itself better, and would make MACAW more
relevant to future networks that are more complicated than just 802.11b.
This archive was generated by hypermail 2.1.6 : Sun Nov 21 2004 - 23:26:02 PST