From: Erika Rice (erice@cs.washington.edu)
Date: Sun Nov 21 2004 - 16:15:00 PST
"MACAW: A Media Access Protocol for Wireless LAN's" by Vaduvur
Bharghavan, Alan Demers, Scott Shenker, and Lixia Zhang:
This paper describes an access protocol for wireless networks. The
protocol, MACAW, is based off an earlier protocol MACA. The protocol is
based on the assumptions they can make for a particular type of radio
technology from Xerox PARC. One weakness of the paper is that the
authors fail to give any motivation as to why this technology that they
are basing their research on is of any importance.
MACAW does make several improvements on MACA. Fairness and utilization
in the face of congestion are improved. The big insight with respect to
congestion is that congestion in a wireless environment with many access
points is not homogeneous. One access point may be congested because
there are many devices within range trying to send, but another may not
be congested because there are few devices in its range. A device in
range of both should be able to tell that the congestion is different.
Another insight of the paper that covers more than their particular
technology is that to ensure reliability of transmission, it is
necessary to have some sort of functionality at the link layer. They
choose to use link layer ACKs. The reason for this apparent violation
of the end-to-end argument is that waiting for a higher layer to time
out takes too long because higher layers need to be more generic. The
link layer can respond quickly because it can be customized to the link.
Thus, the apparent violation is actually perfectly in line with the
end-to-end argument; higher layers cannot achieve the functionality
required.
One final criticism is that while the paper does show significant
improvement over the MACA protocol, it seems like the improvements are
based only on particular problems which manifest themselves in small
networks. One wonders if this focus on the small problems might lead
the authors to miss the fact that some of these problems might be
symptoms of a more fundamental problem which they are not addressing.
This archive was generated by hypermail 2.1.6 : Sun Nov 21 2004 - 16:15:01 PST