Practical Capacity Benchmarking for Wireless Networks

Recent advances in the optimal design of cross-layer congestion control, routing and scheduling algorithms allow us to calculate wireless network capacity [3]. However, for large networks, the algorithm is computationally intractable. In this paper, we provide an upper bound on the wireless network capacity that is computationally efficient to implement. Our upper bound calculations consider multiple different wireless interference models. Our approach draws on results from multicommodity flow over the sparsest cut. We develop a polynomial time randomized algorithm to approximate the sparsest cut in general wireless networks. To ascertain the performance of our upper bound, we use the cross-layer approach in [3] to develop a lower bound by considering only a subset of independent sets. The lower bound is shown to be within 95 percent of our upper bound on average for the primary interference model. For 802.11, 802.16 and the 2-hop interference models, the lower bound is within 70 percent of the upper bound. By applying the sparsest cut recursively in the network, we also develop a capacity heat map that allows us to visualize the regional capacity and identify network bottlenecks.