On Data Source Selection For P2P Based VoD System

With rapid development of Internet technology, more and more network applications access to our daily lives. We could obtain common files, images or multimedia files from Internet conveniently, and especially, we prefer downloading various high-definition video from kinds of website. But unfortunately, the downloading time is usually so long that we can not watch the video until it becomes a complete file stored in local storage, which is a big restriction for promoting such multimedia web applications. In order to make up this deficiency, many video-on-demand (VoD) systems are emerged to improve the quality of service for web client users. During the early days, client-server based architectures are used to support VoD application. With the rapidly increasing number of client users, the video servers can not afford such service pressure. So, peer-to-peer (P2P) technologies are adopted in VoD applications to improve the performance of video servers and the utilization of network resources. BitTorrent (BT) protocol is widely deployed in many P2P applications, and it could also be used for VoD applications. However, some algorithms of the original protocol are not quite suitable for VoD applications. This paper aims to improve these algorithms, and implements a new integrated BT based VoD system.This paper at first describes the characteristics of VoD applications, and analyzes some VoD systems transmission technology. The possibility of combining BT and VoD is also explained. Then based on the detail analysis of functional behaviors of the BT protocol, an improved piece selection algorithm and a revised peer selection algorithm are proposed to support faster VoD application . Besides above advantages, the new solutions could conducive to the rapid access to video files and enhance the overall performance of the VoD system. At last, we implement a BT based VoD system, integrating two algorithms this paper proposed, and carry out certain performance testing on it. According to the testing results, the piece selection algorithm and the peer selection algorithm fully meet the requirements of video-on-demand, such as a short start delay, fewer pause, low level of rely on server, finally improved user experience.