| In recent years, continuous video has emerged as an important element of the multimedia technology revolution. Applications such as digital libraries, distance learning, digital broadcasting and virtual conferencing promise to stretch the information technology to its limits. It is widely accepted that the problem of supporting continuous video (either live or stored) with Quality of Service (QoS) will decide the future course of these important applications, and hence that of information technology itself. A major challenge in this problem is providing tiered (involving multiple, disparate users) QoS while maximizing resources utilization, so that each of the applications can be economically offered.; Providing QoS in an application server, such as a video server, involves admission control and scheduling. Admission control is a means of deciding whom to admit and what service to offer to maximize the available profit. Given any suitable admission control scheme, it is the task of scheduling to ensure that individual users are provided with acceptable QoS. In this thesis, we present a model and a scheduling algorithm which together provide tiered QoS in continuous video applications. The model captures application QoS requirements in a form suitable for scheduling, and the algorithm performs resource management, such that the individual applications are provided with acceptable QoS. The implementation of the algorithm in a prototype video server and its performance involving MPEG encoded video clips and multiple client streams will be presented. The results indicate that the algorithm, (1) ably meets the disparate QoS requirements of users; (2) simultaneously maximizes the server's ability to support a large number of users, and (3) imposes very little implementation overhead. The proposed model and the algorithm are easily applicable to bundled continuous media, including voice and video.; In addition, we present policies for supporting end-to-end deadlines in live video applications. We augment these policies by extending the proposed algorithm to provision multi-tier QoS in a video bridge that facilitates live video teleconferencing involving multiple streams. Simulation results indicate that our approach drastically improves the deadline miss rate and the supportable load for any given deadline miss rate. |
