In this paper, the design issues on a scaleable popularity independent video on demand (VOD) server which has been designed and implemented at Ajou University are discussed. The server with parallel architecture aims to resolve a basic performance bottleneck, which has been encountered in the design steps of many other VOD servers. The performance degradation caused by the skew in video popularity becomes minimized through the realization of server-level striping. The server-level striping increases the scalability of server: the increase in performance follows closely the increase in the number of video pumps. The communication and processing power overhead introduced by the parallel server architecture is relieved by a specially designed meshed pipeline type interconnection network. The design issues on the interconnection network and data placement policy are also addressed with empirical data. In the last part of this paper, the real-time operating system and the schedulers for the video pump are described.
展开▼