On-demand OFDMA: Control, fairness and non-cooperation

Motivated by recent work on improving the efficiency of the IEEE 802.11 protocol at high speeds, we consider an OFDMA system in which the users make reservations requests over a collision channel. The controller schedules from only amongst the successful requests using an alpha-fair scheduler that balances the network throughput and fairness to nodes. We first analyze the performance of the alpha-fair scheduler when used with an Aloha reservation channel. We then assume that the network prescribes reservation rates to active nodes but that nodes may attempt reservations more aggressively so as to be scheduled more frequently (and unfairly). A simple game theoretic analysis of interaction between the Aloha reservation channel and the scheduler shows that in the presence of other cooperative users, a node attempting at a rate higher than that prescribed indeed obtains a larger (unfair) throughput. For such a network we propose a robust alpha-fair scheduler that penalizes aggressive users. This scheduler along with the prescribed reservation rates forms a Nash equilibrium.