Design of a scalable video on demand architecture

Abstract

The SAVoD (Scalable Architecture for Video on Demand) approach to video on demand exploits the fact that end-user latency goals are relatively modest, while increasing bandwidth is much easier than improving bandwidth. Instead of classic approaches which add latencies, making system-level latency goals hard to achieve, the SAVoD approach shares information across multiple users. Unlike other video on demand (VoD) approaches, SAVoD broadcasts a fixed number of replications of a given stream. VCR-like operations are implemented by switching streams (to another instance of the same stream or to a different stream). The result is a system that scales up well: additional network traffic is only generated by an increase in either the video standard or the number of simultaneously available movies, and the number of users has no impact on system requirements. Instead, increasing users makes an improvement in quality of service more affordable. With a modest budget of 32Kbit/s per user, 69,000 users are required to justify a new HDTV movie, or 23,000 users for a broadcast-quality service.