UC San Diego

The object of interest in this dissertation is a cellular wireless system with cooperation among base stations. We study such systems from a cross-layer point of view. In the first part of the dissertation, we investigate the maximum throughput of such a system. Assuming that the relative traffic of each mobile is specified in advance and some simplifying assumptions on the underlying channel, we show that the maximum stable throughput can be expressed in terms of the capacity of the channel. We then formulate a queueing model for this system and propose a throughput-achieving service policy. We then propose a fixed-point approximation as a tool for the performance analysis of this policy. We then proceed, via simulations, to demonstrate the advantage of cooperation over the traditional operation of such systems. Since the proposed policy is computationally expensive, we propose a practical, albeit suboptimal, scheme for large systems. We quantify, again by means of simulations, the advantage of the proposed scheme over the traditional operation. We next study the performance of the queueing network in heavy traffic. Specifically, we prove limit theorems justifying a diffusion approximation for a heavily loaded system operating under the policy proposed earlier. We first show that for a two-user system,the renormalized queuelength process converges in distribution to a semimartingale Reflecting Brownian Motion (SRBM) living in a two-dimensional quadrant. Using different techniques, we next show that for an arbitrary sized system, the renormalized workload process converges in distribution to an SRBM living in the N-dimensional positive orthant where N is the number of users in the system