From: Danny Wyatt (danny@cs.washington.edu)
Date: Mon Nov 22 2004 - 01:44:22 PST
MACAW: A Media Access Protocol for Wireless LAN's
Vaduvur Bhargavan, Alan Demers, Scott Shenker, Lixia Zhang
This paper presents extensions to an existing wireless media access
protocol (MACA) while also providing a good summary of some of the
unique needs of a wireless link layer. Simply sensing the carrier is
not enough for wireless media, since all hosts cannot hear all other
hosts traffic (the "hidden terminal" scenario) and not all traffic will
collide with all other traffic, so medium usage can be misleading (the
"exposed terminal" scenario). MACA handles these scenarios with a
two-way RTS-CTS "announcement" before sending real data. Since the
announcement is two-way, only valid potential sources of interference
will hear both sides of it and back off their transmissions. MACAW
extends the RTS-CTS scheme to also advertise a current back off value so
that a unified metric of the congestion in an are is actively broadcast
instead of being independently inferred by each host. MACAW further
extends MACA by adding an explicit link level ACK so that hosts can more
quickly respond to lost transmissions (MACA left this to the
comparatively slow transport layer); an additional message in the
pre-send announcement (making it a three-way) so that other hosts can
more accurately infer when next to transmit; and an RRTS message so that
receivers can prod senders into faster retransmits to prevent themselves
from being starved out.
Initially, I was skeptical of the extremely simplified "collision-only"
network model on which the paper based its developments: no multipath
problems, and no overlap between base station cells (thus also ruling
out capture and interference). I expected that they would explain a few
corner cases where the standard technique fails and then show that their
extensions outperform. And while they do that to some extent
(especially in the evaluation section), they also outline a few failure
scenarios that re possible under their simplification and for which they
have no solution. That impressed upon me the true difficulty of the
problem. An additional assumption that they don't make explicit but
that is obvious today is the assumption that all hosts are benign. As
described, their protocol is vulnerable not just to denial-of-service
attacks, but to outright exploitation by a greedy host. Altogether, I
am interested to learn how these issues have been resolved (or not!) in
contemporary wireless networks.
This archive was generated by hypermail 2.1.6 : Mon Nov 22 2004 - 01:44:29 PST