This paper explains the design of MACAW, a wireless media-access 
protocol based on MACA, by presenting a linear progression through the 
reasoning and motivations behind each of its features.  The protocol is 
designed, tested and evaluated.  MACAW is derived from MACA by adding 
new control signals and by changing some of the algorithms (back-off, 
congestion accounting on both endpoints, etc).  The authors state their 
goal to be the development of a wireless media access protocol and the 
exploration of some of the basic performance and design issues inherent 
in these protocols.  Additionally, the goal of MACAW is to achieve both 
a high overall throughput and a fair allocation of throughput.
 
The authors set out 4 design guidelines to guide them in the design of 
their new protocol:
 
-- the relevant congestion is at the receiver end, not the sender 
(compare to CSMA)
-- congestion is location dependent
-- to allocate media access fairly, learning about congestion levels 
must be a collective effort
-- the media access protocol should propagate synchronization 
information about contention periods so that contention is "fair" for 
all devices
 
The design of MACAW is documented in a story-line fashion, with each 
change motivated by a compelling example and confirmed within a network 
simulator:
 
-- the back-off algorithm is changed from exponential because of 
unfairness, instead the current back-off counter is shared among devices 
via the packet header and the back-off is done by multiplying to 
increase and subtracting to decrease (MILD)
-- bandwidth should be divided evenly across streams not hosts by 
implementing separate queues in a device for each stream
-- ACK is added to improver performance with link-layer recovery
-- DS is added to indicate that a RTS/CTS exchange succeeded and that 
data transmission is to follow
-- RRTS is added for situations in which a device cannot respond to an 
RTS because of deferral
-- a separate back-off value is necessary for each station and the 
back-off values of both ends need to be inserted into the packet header
 
Several issues remain unresolved such as implementing multicast 
communication and precisely defining fairness.  Additionally several 
enhancements are left un-investigated: piggy-packing ACK on CTS packets, 
NACKs instead of ACKs, full carrier-sense instead of DS packets, power 
variation, etc.  MACAW outperforms MACA in fairness and throughput 
comparisons in the presence of congestion and noise, and performs nearly 
as well in the absence of both.   Despite these performance 
improvements, the authors conclude that their design is still preliminary.
Received on Fri Sep 22 2006 - 18:07:02 EDT
This archive was generated by hypermail 2.2.0 : Mon Sep 25 2006 - 23:03:50 EDT