Large-scale P2PVOD system: Focusing on clients

Abstract

Using P2P network to provide video on demand (VOD) service can support large-scale users so as to save the bandwidth cost of servers. As VOD must support VCR characters (drag, pause, randomly choosing watching position, etc.), building a large-scale expansion of high-performance VOD system still faces huge challenges. The paper first presents a P2P based reference model, defines the design object, and analyzes the difficulties. And then taking PPLive demand system as an example, it gives client-based solutions, including disk cache-out strategy, memory cache prefetch strategy, maintaining relations between neighbors, task scheduling, data upload strategies and so on. PPLive-demand systems are putting into practice at present with satisfactory fluency and drag delay. The largest number of simultaneous online users is over 210000, and online video data is over 700. Compared with traditional client-server model, the average bandwidth saving rate of servers is 88% or more, and the bandwidth saving rate of user-number’s the peak value can be up to 93%.