Degree Name

Doctor of Philosophy


Department of Electrical, Computer and Telecommunications Engineering


Interactive video systems will form the basis of a vast range of new telecommunications services set to emerge in the coming years. Interactive video services enable the user to select and control the playback of video objects located on a remote server, in real-time. Telecommunications, storage and compression technologies are maturing and converging rapidly to make interactive video achievable in the wide area. These interactive video services will transform information and entertainment systems in a fundamental way, the true impact of which is difficult to predict.

This thesis presents a unified treatment of some of the issues related to interactive video service provision. Currently, there is much research being conducted on various aspects of these systems. The complex nature of large-scale interactive video systems has, however, often lead to the use of significant simplifying assumptions or isolated treatments of specific sub-problems. While these efforts are valuable, this thesis takes a more holistic approach. A top-down approach is used to perform a cost comparison of an entire network architecture, before considering various aspects of the system in more detail.

As a result of an extensive literature survey and preliminary investigations, several areas are isolated for further consideration. Specifically, the use of disk-array based storage for video servers is considered in detail. Performability analysis is employed to compare various disk-array architectures from a combined performance and reliability perspective. Further, a new packing scheme is introduced for allocating video objects to the various disk arrays within a server. This scheme is shown to give considerable efficiency improvements over existing packing heuristics.

Before transmission of video can commence, a video server must ensure that sufficient resources are available to support the call. Traditional call-admission control (CAC) procedures are shown to lead to high delay variability due to poor load balance. A new CAC scheme is proposed which utilises significant information about server state to improve load balance and reduce delay variability.

The thesis concludes with a methodology suitable for designing large-scale interactive video systems suitable for supporting video-on- demand style applications. The method applies the analytic tools presented earlier in the dissertation to provide the engineer with a robust method for top-down system design.

The results of this thesis lead to the conclusion that interactive video systems can be constructed cost-effectively utilising existing storage and networking technologies. The costs are, however, substantial and a high level of market penetration will be required to ensure that such systems are profitable in the medium-term.