Design And Implementation Of P2P Video On Demand System

The integration of Internet and video brings the online video. The network video which has interactivity and entertainment has attracted many people to watch video online. In the traditional C/S structure, server’s computing power, memory and bandwidth can easily cause single point system bottleneck. IP Multicast tries its best to delivery data, lacking of congestion control mechanism. Content Distribution Network has poor scalability and high cost. In peer-to-peer network architecture, resource is stored in each node. Computing and storage are collaboratively completed by all nodes. This not only solves bottleneck problem on single point but also improves system robustness and expansibility.Firstly, we research the development of the online video. Summarizes the advantages and disadvantages of using C/S architecture, IP multicast technology, proxy, CDN architecture, P2 P network to design video on demand system. Then we introduce several classic P2 P network on-demand model. We know the key of designing a good video-on-demand system is data scheduling algorithm and building a low-latency network.Based on peer-to-peer and Content Delivery Network, we propose a hybrid data scheduling mechanism, a node not only can get resource from a fixed server, but also can share resource with other nodes in the network. Meanwhile we use BitTorrent protocol on video on-demand. Node uses the modified BitTorrent protocol to communicate with other nodes. Since BitTorrent protocol always select to download the rarest piece first to increase the diversity of the pieces inside the swarm, it could not support time sensitive content. This paper introduces a sliding window mechanism to download urgent data first.In peer-to-peer network, overcoming the problem associated with the dynamic nature of peers is fundamental challenge of peer-to-peer network. Either node leaves the network intentionally or unintentionally. Peer groups frequently change. The resource is not the same among the members. It is a huge challenge to build overlay network which has small transmission delay to maintain application-level connectivity. This paper propose the concept of autonomous domain for discovering neighbor nodes and selects partners by predicting ability, meanwhile regularly updating the neighbor nodes and partner to optimize overlay network.Finally, this paper describes the design of the system framework, then designs and implements the video on-demand system. This paper tests the system in the LAN, analyses important indicators, such as system delay and continuity of playback. Test results show that the client works well. The video on demand system which is based on hybrid data scheduling algorithm and selection algorithm can afford good quality of service for user.