Towards health of replication in large-scale P2P-VoD systems

Unlike P2P-based live streaming, P2P-based video-on-demand (P2P-VoD) has asynchrony in sharing video content among users. To compensate, each peer is expected to contribute a fixed amount of disk space, which forms a distributed P2P storage system. How to regulate this storage system is undoubtedly one of the most important strategies in P2P-VoD systems. PPLive, a large-scale P2P-VoD system, has made great efforts to optimize replication by improving file replacement algorithm, since developed in the fall of 2007. In this paper, we introduce the replacement algorithms PPLive takes and compare their performance when the user request patterns are in a steady state and when they change sharply. These real-world experiences will provide valuable insights into what a healthy replication is and the future design of replication strategy.