On max-min fairness and scheduling in wireless ad-hoc networks

Abstract

This paper addresses the problem of fairness in wireless ad-hoc networks. Usually, the problem of fairness in wireless ad-hoc networks is addressed in a classic approach inherited from wired networks. The common assumption is that nodes/flows have pre-assigned fair shares. The task becomes then to modify wired networks' fair queueing algorithms to address wireless network nature. Wired networks have efficient means of allocating fiar shares through admission control and additionally, the fair shares remain constant throughout the session duration due to the static nature of the nodes. In ad-hoc networks, it is meaningless to assume statically pre-assigned fair shares, since on one hand, not only the nodes move, but also the routers are mobile, and on the other hand, contention is location dependent, such that in terms of absolute guarantees, fairness would at most mean "avoiding starvation": thus applying rate proportional fiar queueing algorithms is beyond the original goal of such algorithm, viz. flow isolation/protection and bandwidth guarantee. In this work we argue in favor of multi-level scheduling for wireless ad-hoc networks with max-min fair allocation of the fair shares at the lower-most layer (MAC layer). This paper, mainly lays down the theoretical framework by which one can calculate the fair shares that would achieve max-min fairness in an ad-hoc network. We design distributed algorithms that allow each node to determine its max-min per-link fair share in a global ad-hoc network without knowledge of the global topology of the network. The results are then used in conjunction with a novel practical scheduling algorithm for IEEE 802.11 to show how fairness is achieved