A Framework for Dynamic Contention Window Adaptation

Network designs with multichannel and multi-interface capabilities are fast emerging as the new solution to the high interference and low throughput problems faced by wireless mesh networks today. Multichannel capability, ideally, isolates transmissions over different channels and hence interference is expected to be minimized. Multi-interface capabilities aim to solve network partitioning problems by making sure that the wireless network remains connected at all times during operation. Practically, however, there are limitations to performance improvement because of channel contention and adjacent-channel interference effects. Moreover, an intelligent channel allocation algorithm has to make sure that it utilizes all the channels effectively and in such a way as to minimize interference to other wireless nodes and also to balance the amount of contention over different channels. In this thesis, we take another approach to improving the performance of wireless networks from the perspective of providing better fairness. Specifically, we build driver-level support in Net-X, a multichannel, multi-interface system for adjusting the contention window adaptively through knowledge of network topology and the level of channel contention sensed by a node.