End-to-end maxmin fairness in multihop wireless networks: Theory and protocol

To promote commercial deployment of multihop wireless networks, the research/industry communities must develop new theories and protocols for flexible traffic engineering in these networks in order to support diverse user applications. This paper studies an important traffic engineering problem - how to support fair bandwidth allocation among all end-to-end flows in a multihop wireless network - which, in a more precise term, is to achieve the global maxmin fairness objective in bandwidth allocation. There exists no distributed algorithm for this problem in multihop wireless networks using IEEE 802.11 DCF. We have two major contributions. The first contribution is to develop a novel theory that maps the global maxmin objective to four local conditions and prove their equivalence. The second contribution is to design a distributed rate adjustment protocol based on those local conditions to achieve the global maxmin objective through fully distributed operations. Comparing with the prior art, our protocol has a number of advantages. It is designed for the popular IEEE 802.11 DCF. It replaces per-flow queueing with per-destination queueing. It achieves far better fairness (or weighted fairness) among end-to-end flows. (C) 2011 Elsevier Inc. All rights reserved.