From: Craig M Prince (cmprince@cs.washington.edu)
Date: Mon Nov 22 2004 - 04:40:55 PST
Reading Review 11-22-2004
-------------------------
Craig Prince
The paper titled "MACAW: A Media Access Protocol for Wireless LAN's"
described a new approach for dealing with collision detection/avoidance on
wireless systems via media access control. Existing schemes for wired
networks use the fact that a potential transmitter can listen on the line
for other traffic and know if there is interference. This is the Carrier
Sense Multiple Access (CSMA) approach; however, for wireless this is not
an option because 1) transmission distance is finite/variable and 2) your
own transmission is so strong that you can't detect any interference while
you are transmitting. As a result, the authors developed a new protocol
for getting access to the wireless "airwaves" that is a reservation type
system.
The authors built their protocol upon a previously proposed MACA protocol,
which was based on four observations about the wireless environment.
First, that "contention is at the receiver not the sender", so CSMA won't
work. Second, that interference depends highly on location in the network.
Third, that everyone must learn about congestion to be able to provide a
fair allocation. Fourth, that you need to synchronize everyone to be able
to make a good use of the media. From these observations, the key insight
is that you create a scheme where you ask specifically before
transmitting: RTS-CTS.
The authors claim that the MACA protocol is flawed in that it isn't fair
to all clients and they propose a number of changes to make it better. The
first change is in how the nodes backoff after a collision. Instead of
exponential backoff, they employ a multiplicative increase linear decrease
backoff. Also, they add a mechanism to synchronize everyone's backoff
after a transmission. This results in a fair chance for everyone to
transmit. Another improvement is that they add an ACK message directly to
this link layer protocol to avoid retransmissions at the TCP level.
Finally to handle two cases where a node can still dominate they add a DS
and RRTS messages which prevent a single client for using all the
bandwidth.
I liked how this paper addressed the issue of access to wireless media
from the ground-up. Instead of relying on assumption in wired networks,
this paper directly addresses why the existing access protocols don't work
and systematically develops a protocol that will (hopefully) work. Another
thing I liked about this paper is that they clearly state the cases where
they know that their protocol will still fail. By listing the protocol's
limitations they give the reader a better understanding of the protocol.
I had several problem with this paper. I thought that the simulations
conducted were very superficial and seemed to be designed specifically to
target cases where they knew one would fail and the other would succeed.
While this showed that their protocol has some good properties, it did not
convince me that their protocol is necessarily the best one to use. Also,
how their protocol performs in simulation and in practice seem like two
very different things. I was skeptical that their simulations were not
completely realistic. Finally, their Xerox PARC network was different then
most wireless network and I'm not sure whether their proposed protocol
would still be effective in a more traditional wireless environment where
their a lot more interferrence.
Overall, I think we can view this paper as an example of one way to design
new protocol. By carefully analyzing the flaws of existing protocols and
determining why those flaws occur, we can target and fix these
specifically in our protocols.
This archive was generated by hypermail 2.1.6 : Mon Nov 22 2004 - 04:40:55 PST