(no subject)

Gennady Pekhimenko

 Summary.
This paper describes a problem of constructing protocols for wireless
LAN networks. This problem is very actual, because of the variety of
personal digital devices that have already appeared in recent years.
This paper studies media access control protocols for a very specific
type of wireless networks (constructed at Xerox), but these ideas can
be used for many others similar wireless LANs.

As to the authors the main question is how to control the access to
shared media for wireless network. Also, they have two main goals as
to their research :
-- create media access protocol for use in the wireless network
constructed in Xerox Research Center
-- improve performance for the basic existing algorithms.

The designed was based on the four key principles:
-- relevant contention is at the receiver
-- congestion is location independent
-- we need to use collective information
-- we need to propagate synchronization information about contention
periods.

Then we get information about network type (PARC's Radio Network) and
some special assumptions they've made: two base station are always
within the range of each other; possible noise, but research has been
done in a noise-free setting.

Next, paper explains us why they didn't not use CSMA and shows the
main ideas of MACA. The key difference between MACA and CSMA is that
MACA has the RTS-CTS exchange that helps nearby stations to avoid
collisions at the receiver.

Next chapter is about creating their new MACAW protocol that can solve
problems of simple MACA. While analyzing the backoff algorithm we see
that the main problem is that there is no sharing information about
congestion. Also authors propose two more features for improving
performance results:
 -- multiple stream model (allocate bandwidth equally to streams)
 -- basic message exchange changes : ACK (using acknowledgement
 packet); DS (30 byte Data-Sending packet); RRTS
 (Request-for-Request-to-Send packet - trying to solve lack of
 synchronizing information).

At the end, paper describes the issues that still have not been
supported: NACK, good multicast etc.

In general, I think the paper is good. It describes new ideas for
creating wireless LAN protocols and improves existing ones.
But this design is really preliminary and they need to make a lot of
things like asynchronous service, multicasting etc. Another thing I'd
like to mention is that they use some kind of packet-level simulator
(modification of existing for wired ones). So we can only hope that
their
results would be the same for real wireless LANs as well as on
simulator.
Received on Thu Sep 21 2006 - 12:07:28 EDT