Quality of service control for distributed multimedia systems

Shunge Li, Purdue University

Abstract

This thesis centers on the fundamental mechanisms of performing Quality of Service (QoS) control for packet video over an integrated services network such as the Internet and enterprise intranets. It first proposes a bandwidth allocation scheme that combines a bandwidth reservation approach with an adaptation approach to offer service guarantee and bandwidth sharing. It achieves a high degree of bandwidth sharing by taking advantage of the adaptability of RT applications to network band-width and by allowing them to dynamically vary their bandwidth requirements. The dynamic bandwidth contention is resolved using a market mechanism found in microeconomics. Experiments show that this hybrid scheme enables better bandwidth utilization and provides certain degrees of isolation and sharing in a fair manner. This thesis then investigates the adaptability of the video conferencing (VC) application to the run-time network bandwidth variations in the Internet environment. Several techniques are developed to make VC software highly adaptable to the currently available network bandwidth while maintaining a certain level of QoS perceived by end users. Experiments reveal that these techniques are highly flexible and computationally less intensive, and support graceful degradation of video quality. A flexible and programmable facility, called an active gateway, is developed in this thesis to serve as an infrastructure for customized QoS control. In addition to its traditional routing function, an active gateway can recognize and execute Tcl scripts embedded in packets passing through it. This capability brings to distributed systems such benefits as customized QoS control, cooperative computing, active searching, knowledge discovery, network load balancing, and network programmability, some of which are illustrated through experiments. Finally, this thesis presents a unified framework under which adaptive, flexible, and customized QoS control can be performed together.

Degree

Ph.D.

Advisors

Bhargava, Purdue University.

Subject Area

Computer science

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS