Simple fairness metric for time-shared wireless data networks

Third generation networks (e.g, 1xEV-DO, 1xEV-DV and WCDMA) all improve the efficiency of data transmissions by taking advantage of the fact that data is less sensitive to delays than voice. Therefore, instead of serving a data stream synchronously, they allow for bursty servicing of each data stream. In the forward link, each data stream burst is typically scheduled when the corresponding mobile is in favorable radio conditions to maximize radio resource efficiency. This results in a trade-off between total sector throughput and fairness among users within the sector. Total sector throughput is increased by serving users in good average radio conditions while fairness is achieved by providing all users with identical throughputs. Naturally, these two objectives conflict and some trade-off must be made between them. While sector throughput is easily measured, the fairness of different schedulers have typically been compared by visual inspection of the cumulative distribution function of user throughputs. We recommend using a more objective criterion for comparing schedulers, namely the Jain fairness index. We illustrate how this metric accurately reflects fairness and provide examples showing the throughput/fairness trade-off.