Improving the performance of wide area networks

Abstract

Research in to the performance of wide area data networks is described in this thesis. A model of wide area network packet delays is developed and used to direct the research in to methods of improving performance. Wide area networks are slow and expensive compared to the computer systems that rely on them for communication. Typically data networks are packet switched in order to make efficient use of resources. This can lead to contention, and the mechanisms for resolving contention can bring about further delays when demand for resources is high. In this thesis, network users are viewed as interacting decision makers with conflicting interests, and Game Theory is used to analyse the effects users have on each other’s performance. It is asserted in this thesis that wide area network performance is an ethical issue as well as a technical one. Compression is examined as a technique for reducing network traffic load. While load reductions can reduce the time packets spend waiting in buffer queues experimental results show the compression process itself can present a bottleneck if CPU resources are limited. The other inhibiting factor with regard to wide area network performance is the time it takes for a signal to propagate through a transmission medium. Propagation delays are bounded by the speed of light and becomes significant as the distance between computer systems increases. Mirrors and Caches are methods of bringing data closer to the user, thereby reducing propagation delays and capping traffic loads on long haul communication facilities. The performance benefits of replicating data within a wide area network environment are studied in this thesis.