Optimal robust policies for bandwidth allocation and admission control in wireless networks

We consider joint strategies of bandwidth allocation and admission control for elastic users competing for a downlink data channel in a cellular network. For the sake of robustness and generality of the results we focus on the set of strategies whose performance does not depend on the detailed traffic characteristics beyond the traffic intensity. Performance is studied at the flow level in a dynamic setting were users come and go over time. A number of user classes are considered, which are characterized by their achievable bit rate, guaranteed throughput, arrival rate and mean flow size. We aim at characterizing a strategy which is optimal in the sense of having the lowest blocking probability. Such characterization provides some interesting insights into the optimal policy and its evolution as the system load increases. In some cases computing the optimal policy can be exceedingly complex except for lightly loaded systems. For those cases we propose a computationally feasible suboptimal policy that achieves a good relative performance. Finally, we show that in scenarios of practical interest, the loads of interest lie inside the region where the optimal policy can be efficiently computed.