<-- Back to Main Page

Date : Monday, Oct 31, 2005
Speaker : Ananth Rao
Affiliation : UC Berkeley
Talk Title : Achieving End-to-End fairness in wireless networks with randomized TDMA schedules
Slides :

Abstract

The MAC layer plays a very important role in the performance and fairness of multi-hop wireless networks. MAC protocols can be classified into either contention-based or timeslot-based approaches. Both simulations and deployed test-beds show that contention-based MACs can lead to poor-fairness in a multi-hop network due to hidden terminal problems and other asymmetric interactions. Time-division mutliplexed (TDMA) approaches can potentially solve such issues because they explicity disallow competing stations from transmitting at the same time. However, TDMA schemes suffer from two important drawbacks which have slowed their adoption. Firstly, they require an explicit setup and teardown phase for each flow in the system. This overhead can be very significant for web-traffic like workloads with many TCP arrivals and departures. Secondly, TDMA schedules have to be recomputed when link quality changes. The allocation at each hop of a multi-hop flow must be the same, and hence the change in the data rate of any intermediate link will result in the entire schedule being recomputed. In this paper we propose a simple distributed algorithm that addresses these two drawbacks by using an increase-decrease algorithm to assign weights. These weights have local singnificance only, and the algorithm uses information only from directly competing queues to guarantee convergance to a fair allocation. Almost all control messages are piggy-backed on existing traffic to keep the overhead low. We evaulate our algorithm though both simulations in ns2, and an implementation on top of the Overlay MAC Layer (OML) in a small 802.11-based testbed.