Peer-to-peer Network-based Video On Demand System Research And Design

With the development of networking technologies and the popularization of high-bandwidth networking, more and more clients are accustomed to watching video on the Internet. Traditional Client/Server(C/S) centralized video services require high performance and capability of the server, and as the centre which provides all the video data, the server is inclined to become the bottle-neck of the system. Although some back-up servers may be located among the Internet, the cost of system is relatively high. Peer to peer (P2P) networking design can to some extent solve the problem of traditional video service. P2P networking can take advantage on the spare capability of the clients for providing service for more clients. In this way, the service cost for each client can be dramatically reduced. This paper focuses on the problems of Video-on-Demand(VoD) service. From the points of view of constructing an integrated system, this paper introduces a prototype system DCSVS, which takes searching, transferring, storage, performance of video service system as a whole into account. DCSVS is mainly designed for set top box video service, and this paper gives an introduction of the research and design about this system.The architecture of DCSVS is combined of Management Server, Video Server and Distributed Hash-Table(DHT) overlay networking. The DHT overlay networking at the same time has a structure of cluster-based hierarchical mesh. The mesh structure is used for the non-structuring searching designed for relatively small-scale client networking. The DHT is used for the structuring searching designed for relatively large-scale client networking. Non-structuring searching algorithm PVPS (Peer via Peer Searching) employs a content-based clustering mesh structure in organizing nodes, and in each cluster there are several broker nodes managing others. In the searching process between clusters, PVPS uses (1)altered heuristics MPR algorithm, (2)self-pruning based on neighbor priorities and (3)pruning based on message redundancy to optimize the unstructured P2P searching. Structuring searching algorithm uses improved Kademlia DHT protocol, with the traits of set-top box' s networking stability, can achieve high performance in large-scale clients networking.Data transferring is one of the key problems in video service, and DCSVS employs several policies including chunking based on accessing probability, push-pull combinative transferring, provider selecting algorithm and so on. In order to improve the service performance and to support VCR operations, we introduce anchors into this system and use several diffuse strategies. In order to solve the issues of free-rider and hot-spot, DCSVS uses effective stimulating policies and load-balancing policies. Some of the data played before may be stored in the clients for re-usage, and there are some storage and buffer management strategies. Moreover DCSVS uses effective recovery algorithm to relieve the influences of clients' leaving and networking' s deterioration. Both theoretical analysis and experimental results show that strategies in DCSVS perform greatly in terms of efficiency and robustness, and could maintain a fine state and scalability in large-scale networking.