MACAW: a media access protocol for wireless LAN's

Motivation: This paper studies the MAC protocols for a single channel
Wireless LAN and propose a new protocol ( MACAW ) by introducing
improvements in the existing protocol MACA. The design philosophy is
based on ensuring high network utilization and fair access to the media
(fairness being given higher priority).

Key Points:

1) The paper develops on the following :
a) Relevant congestion control is at the receiver and not the sender.
b) Congestion is not a homogeneous phenomenon.
c) Congestion information should be a collective enterprise.
d) MAC should propogate synchronization information about contention
periods.

2) Backoff counters have been shared across nodes to ensure fairness
while contention as well as MILD has been proposed to avoid repeated
escalation in backoff counters after each successful transmission. The
reason for choosing the numbers 1.5 and 1 is unclear. This results in a
Backoff Algorithm which increases faster than it decreases and
subsequently leads to unfairness (pointed out later in the paper). I
could not understand the motivation for faster increase than decrease.
(They only modify it for the sake of boundness which doesnt justify the
need to make it increase faster)
3) Per stream backoff counter has been suggested to ensure fairness in a
multi-stream model. I really liked the differentiation between per-node
and per-stream contention which I had before overlooked.
4) Addition of ACK mechanism (recovery at link layer) for better throughput.
5) Out of the two approaches suggested (CSMA/CA vs DS) the advantage of
DS seems to be that it doesnt require constant carrier sensing. This
could be beneficial for power constrained devices. However, an adequate
comparison of both the techniques seems to be mising in the paper.
6) The motivation for introducing RRTS and not using just CTS is unclear
in the paper. One reason could be that for RRTS people need to wait till
the next CTS but for CTS they need to wait for DATA packet transfer time.
7) The problems with the backoff counter copying algorithm has been
outlined in the paper. It could lead to the belief of false congestion
in the network. Thus, using a distibuted algorithm for contention and a
single ambient value which gets shared across all nodes seems to be
causing problems.
8) Also, the setups used by them are overly simplistic (ignoring capture
and interference) and narrow in context (indoor wireless LAN) which
would be one of the basic problems in terms of scalability of the protocol.

One of the major contributions of the paper is its thorough
investigation and analysis of problems in the existing protocol(MACA).
They not only raise the problems, but also for the most part suggest
solutions and show that they do work in practice. However, they have
left some questions unanswered and have provided the motivation to dig
in them deeper.

Lesson Learnt: Carefully consider all the different cases while
proposing a new idea.
Received on Thu Sep 21 2006 - 11:02:03 EDT