Channel capacity fair queueing in wireless networks: Issues and a new algorithm

Abstract

Wireless fair queueing algorithms have been extensively studied recently. However, a major drawback in existing approaches is that the channel model is overly simplified - a two states (good or bad) channel is assumed. While it is relatively easy to analyze the system using such a simple model, the algorithms so designed are of a limited applicability in a practical environment, in which the level of burst errors are time-varying and can be exploited by using channel adaptive coding and modulation techniques. In this paper, we first argue that the existing algorithms cannot cater for a more realistic channel model and the traditional notion of fairness is not suitable. We then propose a new notion of fairness, which bounds the actual throughput normalized by channel capacity of any two sessions. Using the new fairness definition, we propose a new fair queueing algorithm called CAFQ (Channel Adaptive Fair Queueing), which, as indicated in our numerical studies, outperforms other algorithms in terms of overall system throughput and fairness among error prone sessions.