Modelling Throughput and Starvation in 802.11 Wireless Networks with Multiple Flows

Carrier sense multiple access (CSMA) protocols require stations that wish to transmit to first check the shared medium for ongoing transmissions. However, in wireless networks stations may not be able to sense transmissions from all other stations; hence those at different locations can experience different amounts of contention, resulting in unfairness and possible "starvation" (inability to acquire channel access for long periods). In this paper we model how the 802.11 MAC protocol allocates bandwidth among multiple saturated flows in a linear topology when not all senders are within sensing range. We consider a specific class of topologies consisting of n independent flows, where each sender can sense k neighbors on either side. Our work uncovers global interactions among flows leading to startling sensitivities in node throughput. A new model to predict the long-term throughput of each flow under saturation is presented and our model results validated via OPNET simulations.