Abstracts Computer Science

Improving System Performance for Wireless Networks

by Fangyang Shen

In this dissertation, we solve two problems in wireless networks: Transmission Control Protocol (TCP) enhancement and cluster node sleep scheduling. When TCP is deployed in a wireless mesh network, it encounters several challenges: transmission errors, packet reordering due to multipath routing, interference among multihop wireless links, and congestion in both wired and wireless networks. Most of the existing TCP enhancements do not properly treat all four challenges. In this dissertation, we propose a new TCP enhancement method,called Congestion Coherence, which can effectively decouple congestion signals from different types of packet recovery. This allows all four aforementioned challenges to be handled properly in wireless mesh networks. Simulation results show that this method improves TCP performance significantly. In cluster-based sensor networks, part of the sensor nodes can be switched into a sleep state in order to conserve energy, if their neighbors can provide the same or almost the same sensing coverage. However, as the number of nodes in the sleep state increases, coverage for the cluster is degraded. It is crucial to maintain high coverage of clusters in order to preserve performance. In this dissertation, we propose a coverage-aware sleep scheduling (CS) algorithm to improve the coverage of each cluster. Compared with two previous schemes: the randomized scheduling (RS) scheme and the distance-based scheduling (DS) scheme, the CS algorithm maintains higher coverage, while guaranteeing the same lifetime for the cluster. The CS algorithm thus improves the overall performance of the cluster-based sensor networks.