Methods of Congestion Control for Adaptive Continuous Media

Since the first exchange of data between machines in different locations in early 1960s, computer networks have grown exponentially with millions of people now using the Internet. With this, there has also been a rapid increase in different kinds of services offered over the World Wide Web from simple e-mails to streaming video. It is generally accepted that the commonly used protocol suite TCP/IP alone is not adequate for a number of modern applications with high bandwidth and minimal delay requirements. Many technologies are emerging such as IPv6, Diffserv, Intserv etc, which aim to replace the onesize-fits-all approach of the current lPv4. There is a consensus that the networks will have to be capable of multi-service and will have to isolate different classes of traffic through bandwidth partitioning such that, for example, low priority best-effort traffic does not cause delay for high priority video traffic. However, this research identifies that even within a class there may be delays or losses due to congestion and the problem will require different solutions in different classes. The focus of this research is on the requirements of the adaptive continuous media class. These are traffic flows that require a good Quality of Service but are also able to adapt to the network conditions by accepting some degradation in quality. It is potentially the most flexible traffic class and therefore, one of the most useful types for an increasing number of applications. This thesis discusses the QoS requirements of adaptive continuous media and identifies an ideal feedback based control system that would be suitable for this class. A number of current methods of congestion control have been investigated and two methods that have been shown to be successful with data traffic have been evaluated to ascertain if they could be adapted for adaptive continuous media. A novel method of control based on percentile monitoring of the queue occupancy is then proposed and developed. Simulation results demonstrate that the percentile monitoring based method is more appropriate to this type of flow. The problem of congestion control at aggregating nodes of the network hierarchy, where thousands of adaptive flows may be aggregated to a single flow, is then considered. A unique method of pricing mean and variance is developed such that each individual flow is charged fairly for its contribution to the congestion.

Grantor dc:publisher

Oxford Brookes University

Year dc:date

2002

Author dc:creator

• Tater, Shalini

Contributors dc:contributor

• The Author
• Ball, Frank
• Kirkby, Paul

Statement dc:rights

• All rights reserved

Language dc:language

en

DOI dc:identifier

https://doi.org/10.24384/xymy-5516

OAI identifier oai:identifier

tle:78d7abac-e358-49e3-a787-c6d2e6213c6a:d6bd9758-527a-46cd-bfe2-c433766e8fca:1