Source-Differentiating Rate-Based Congestion Control Scheme for High Speed Communication Networks

Ketil Danielsen, Taieb F. Znati · D-Scholarship@Pitt (University of Pittsburgh) · 1991

The advent of high speed communication networks created new opportunities for real-time applications. The support of these applications require the development of adequate communication protocols. Several issues related to the transport of multimedia communications are currently being reassessed. Among these issues, the problem of congestion and flow control in packet switched networks is seeing renewed interest. The task of flow and congestion control is to regulate the amount of data admitted to the network in such a way that resources such as logical channels, buffers, transmission bandwidth, and processor time are efficiently utilized and user performance requirements are met. Congestion in an internet environment, consisting of numerous local and wide area networks connected by bridges and routers, is particularly challenging. The current trend at the network design level specify a new channel abstraction to efficiently support guaranteed levels of performance for a variety of applications. In such an environment congestion control may seem to be easier to exercise. However, the characteristics of the real time applications vary dramatically, and are anticipated to stress correspondingly different designs issues on congestion control protocols. Furthermore, the presence of the "best effort" traffic will still require some form of network control. Consequently, solution falling outside the traditional datagram and virtual circuit paradigms must be provided. We propose a hybrid, adaptive congestion control mechanism based on a feedback control to regulate traffic with different requirements. Based on the requirements specified by the application, the scheme allows the network to selectively throttle the channels with the less stringent real-time requirements in order to alleviate congestion.

Read the paper · More papers on PaperTik