Achieving application level fairness through utility-based wireless fair scheduling

Abstract

The increasing number of wireless data users and the deployment of broadband wireless networks have brought the issue of providing fair channel access to the forefront. Wireless fair scheduling algorithms must deal with unique domain-specific issues such as location-dependent and bursty channel error. In the past, a number of scheduling disciplines have been proposed for approximating fluid fair queueing in wireless channels, but using only the ARQ-based link-layer retransmit model. In this work, we generalize past work in three key ways: (a) we use the notion of fair scheduling, wherein the goal of the scheduler is to serve flows in a way that maximizes the aggregate utility of the system, (b) we combine the techniques of FEC and ARQ in a generalized framework in which different scheduling disciplines can be realized by mixing and matching component algorithms within the framework, and (c) we support the notion of level fairness when flows see different channel error probabilities. Specifically, we explore the combination of adaptive FEC and ARQ-based schemes within the utility-based wireless fair scheduling paradigm. We show through simulation and analysis that this combination enables the scheduling disciplines to optimize the aggregate system utility and achieve application level fairness, while handling varying channel errors effectively.