Design and analysis of wireless and wireless -wireline systems

In recent years, we have seen tremendous growth and proliferation of wireless systems. These systems have now become ubiquitous and in turn are spawning new businesses and technologies that provide exciting new applications involving sensing, computing, and communications. This thesis studies the fundamental performance limits of such systems and addresses challenges that arise in their design and analysis. We first consider single-hop wireless systems such as wireless LANs. A key challenge in such systems is to estimate the throughput that an end-user device can obtain using a standard multi-access protocol such as IEEE 802.11 DCF. We propose a mathematically rigorous technique for estimating the saturation throughput of each individual node in such systems. Next, we consider static multi-hop wireless systems such as mesh networks. We focus on the cross-layer design of such systems with specific emphasis on the design of joint congestion control and distributed scheduling schemes with provably good performance. We show that the optimal scheduling problem in such systems is often NP-Complete and design low-complexity distributed schemes with constant-factor throughput guarantees under realistic interference models. We then consider mobile multi-hop wireless systems with many-to-many communication paradigm. It has been shown that mobility can improve the throughput capacity of such systems at the cost of delays. We establish the existence of fundamental trade-offs between the throughput capacity and delay and show that these trade-offs depend strongly on the nature of node mobility. In particular, the delay that must be tolerated in order to improve the throughput capacity with the use of node mobility can significantly differ across various mobility models. Finally, we consider multi-hop wireless systems with many-to-one communication paradigm that is prevalent in certain classes of sensor and mesh networks. Multi-hop routing and many-to-one communication paradigm lead to a severe disparity in the relaying burdens of nodes that can result in the formation of bottlenecks and partitions, limiting the throughput capacity and lifetime, respectively, of such systems. We show that both average relaying burden and the disparity in the relaying burden can be significantly reduced with the addition of a limited infrastructure to such systems