Congestion control in integrated service networks

In this dissertation we study the design of adaptive decentralized control schemes to control the traffic characteristics of connections. We examine how the arrival rate of traffic at network switches can be adjusted using explicit feedback based congestion control mechanisms. We investigate how such modulation may be achieved for three classes of applications, namely data, image and video. We show that the use of predictive control mechanisms can be used to effectively counter the effects of the relatively large time delays in high speed networks. We study the performance of the control mechanisms both in terms of the utilization of network resources as well as the service quality seen by users. Our results demonstrate that well designed, feedback-based congestion control mechanisms perform well in high speed integrated service networks. In this dissertation we also investigate how scheduling mechanisms at network switches may be used to allocate bandwidth dynamically between different classes of traffic. We show that the use of fairly simple techniques can improve performance when the service quality requirements of different traffic classes differ significantly.