Interactive video-on-demand systems in multicast environment

Abstract

Multicast delivery is one of the feasible solutions for the development of a large-scale interactive Video-on-Demand (VoD) system. The major problems of a multicast VoD system are its large start-up latency and difficulties in supporting interactive functions. This thesis presents the results of our work on efficient implementation of an interactive VoD system in multicast environment. A number of contributions have been made in enhancement and analysis of the provision of true on-demand services with full VCR functionality such as Pause (PAU), Fast Forward (FF) and Fast Rewind (FR) in a multicast VoD system. In order to minimize the start-up latency of a multicast VoD system, a new batching technique called double-rate (DR) batching is proposed. The DR batching tries to merge the customer requests on the fly by the use of buffering. Both analytical and simulation results show that the start-up delay is insensitive and over 90% bandwidth reduction can be achieved comparing with the unicast system. A method called Single-Rate Multicast Double-Rate Unicast (SRMDRU) is then developed to support interactive operations in the system by the use of receiver buffer. A performance model is also derived to evaluate the effectiveness of the SRMDRU system. The results show that our method achieves a better performance in all aspects of parameters when comparing with other multicast VoD systems. As there are over hundreds of movies in a large-scale VoD system, it is obvious that no robust delivery scheme can efficiently serve all videos with difterent popularity. Although the benefit from multicast delivery is obvious in terms of bandwidth requirement and scalability, broadcast transmission may be more efficient for those very popular videos having a very high request rate. Also for those unpopular videos, multicast transmission will not gain any advantage and thus a simple point-to-point transmission is preferred. Based on this idea, a novel delivery strategy is proposed for the VoD system to allocate resources according to the popularity of video. We first extend the idea of the DR batching to develop the so-called Slotted Double-Rate (S_DR) batching scheme for serving the most popular videos. A novel algorithm is then derived to select the appropriate delivery method for different videos so that not only the overall bandwidth of the system is minimized but also a preset quality of service is provided for each customer. The result shows that a significant bandwidth reduction can be achieved when using our proposed delivery strategy. Finally, an adaptive batching scheme is proposed to further improve the system throughput. As the performance of the DR batching policy varies with the batching time, we relax the opening time of the multicast streams and dynamically determine the batching time with the average arrival rate given. A new scheduling policy called Least Double-Rate (LDR) is also developed for the order of choosing video for showing when resource is available. Other issues like buffer reservation and VCR resume delay are also investigated. The result shows that the system performance is significantly enhanced by using our adaptive scheme. With the outcome of our investigations, a completed solution for cost-effective implementation of a large-scale interactive VoD system in multicast environment can be derived.