A heterogeneous approach to fair resource allocation and its application in femtocell networks

Abstract

We have recently extended the notion of alpha fairness to include time scale considerations for fair assignment of resources. Two extreme cases are elastic traffic (file transfer) and interactive voice. In this paper we consider time scale separation in the fairness that is related to the mobility of the users. We propose and solve the problem when the utilities are linear in the resources. We apply these results in the context of fair resource allocation in femtocell networks in a dynamic setting. We show how mobility and the constraints on the averaging durations impact the amount of resources each user gets. I. INTRODUCTIONResource allocation algorithms often try to achieve fairness and efficiency. In particular, Opportunistic scheduling algorithms in both uplink and downlink achieve higher throughputs by giving preference to mobiles with better relative channel conditions. Preference according to relative channel conditions rather than absolute channel conditions mean that the radio conditions of each mobile are normalized by the averaged conditions of that channel until then, and mobiles with best normalized radio conditions are selected for transmission; this guaranties fairness.Averaging the radio conditions is done in practice using some low pass filter. Thus averaging is done over some effective period . A longer achieves a larger opportunistic gain at the cost of longer starvations periods. In other words, efficiency is obtained at the cost of being more unfair over a short time scale. Where as elastic traffic may prefer to be insensitive to short time scale unfairness, interactive real time applications may need averaging over shorter time.In a recent work, we introduced the concept of -scale and Multiscale fairness [1] . These new concepts allow distribution of network resources fairly among different classes of traffic and end-user applications. This work mainly focused on the notion of fairness over time and illustrated how resources are fair shared by considering some interesting applications.In [1], we considered abstract time varying channels, time scales were related to the channel coherence times. In the present context, we assume that user mobility is the source of channel variations and the timescale of fairness is determined by spatio-temporal behavior of the user. For example, such situations arise when resource allocation has to take into account that an user can be located at any random location over his region of mobility. The fairness concepts developed with this notion can be applied for example in small cell networks (ex. pico or femto cells). The scheduler in this case takes into account the expected utility of the user over his region of mobility to compute the resource allocation.