Review: MACAW:A Media Access Protocol for Wireless LAN

The paper develops a new meda access protocol, MACAW, for wireless LAN
based upon the MACA protocol. The authors observed that contention is at
the receiver (not the transmitter), congestion is location dependent,
learning about congestion levels should be collective to allow fair
access of the shared media, and finally that synchronization information
about contention periods should be propogated so that all devices can
contend effectively. With these observations in mind, and with the goal
of delivering high network utilization as well as fair access to the
media, the paper discusses changing the original MACA RTS-CTS-DATA
protocol into a MACAW RTS-CTS-DS-DATA-ACK protocol.
The protocol changes summarized:
1) Replace the binary exponential backoff (BEB) algorithm with
multiplicative increase and linear decrease (MILD) backoff algorithm.
They argue MILD backoff is more efficient than the BEB since it prevents
the backoff counter from varying widely.
2) Modify the backoff algorithm so that the current backoff counter is
included in the packet header, thereby distributing congestion
information to provide more fair contention.
3) Make separate queues (with separate backoff counters) for each stream
at a station in order to distribute bandwidth more fairly amongst the
streams (rather than amongst stations).
4) Add an ACK after data transmission to improve efficiency of reliable
data transfer since the TCP timeouts are long and result in long waiting
periods.
5) Add a DS (data-sending) packet sent by the transmitter after it
receives the CTS from the receiver and before it sends the data. This
informs other stations that RTS-CTS was successful and the DATA-ACK is
about to follow. This synchronization allows other stations to defer
transmission till after the ACK is received and hence reduces
collisions.
5) Add an RRTS if needed. This also provides synchronization in the case
that a transmitter is sending RTS and the reciever is getting it but has
to defer transmission of a CTS to prevent collision with another
transmitter. When the receiver is free to receive and transmit, it sends
an RRTS to inform the transmitter that it may start the
RTS-CTS-DS-DATA-ACK cycle.
The authors eventually show that despite the modifications to message
sending protocol there are still cases where one transmitter can gain
complete control over the channel and starve another transmitter. They
also show that the changes to the backoff algorithm can cause problems
in non-homogenous congestion through leakage of the single backoff
counter from one cell to another that have different congestion levels
leading to an inaccurate view of contention. Also, it aggravates the
problem when there is no contention but either the RTS or CTS is
corrupted due to noise, because the backoff increases even though there
is no congestion.
The paper evaluates two scenarios and shows that despite the increase in
overhead, there is an increase in throughput due to better congestion
handling (assuming offcourse that there is congestion). The experimental
results also show fairer division of throughput among streams in the
same cell and better handling of non-homogenous congestion (although
they previously mentioned that non-homogenous congestion was a
problem).
I find the paper explains the protocols well, however, they seem to go
on adding overhead to handle specific cases and eventually still end up
with unsolved problems. I am unconvinced that MACAW is much better than
MACA.
Received on Mon Sep 25 2006 - 09:47:55 EDT