On the fairness delay trade-off in wireless packet scheduling

We consider the problem of downlink packet scheduling in a time-slotted wireless communication system when a hybrid automatic repeat reQuest (H-ARQ) re-transmission strategy is adopted. User level fairness and average delay per packet are two important metrics used for evaluating the performance of a scheduling policy. However, these are two competing objectives. A good scheduling policy must be flexible enough so that it can be tuned to trade-off one objective for another. We propose one such scheduling policy characterized by a single parameter that can be varied to capture points on the trade-off curve. Our approach is to study two reduced versions of the original problem and construct delay and fairness optimal scheduling policies based on the optimal solutions to these reduced problems. We then leverage intuition from these optimal policies to design a heuristic scheduler that captures the fairness v/s delay trade-off. We demonstrate the efficacy of the proposed scheduler over benchmark schedulers like round-robin (RR), maximum SNR or C/I and proportional fair (PF) via link level simulations. The proposed policy offers a superior fairness and delay performance and is also low complexity from an implementation perspective.