节点文献

基于P2P的视频点播系统的应用研究

Applied Research on P2P-Based Video-on-Demand System

【作者】 万川龙

【导师】 桑军;

【作者基本信息】 重庆大学 , 计算机软件与理论, 2008, 硕士

【摘要】 视频点播(Video-on-Demand,简称VoD)是以用户需求为主导的视频系统。如何提高视频点播系统的可扩展性和在动态环境中的可靠性,成为视频点播系统大规模应用的关键。传统的视频点播系统基于C/S(Client/Server)模式,它为每个请求分配一条数据流,当用户数量大规模增长,特别是点播流行节目时,系统将难以提供质量保证,因而无法满足大量的点播服务。为此,研究者提出了基于P2P的流媒体分发技术。P2P流媒体通过利用普通节点(Peer节点)的闲散资源(带宽、CPU以及存储资源等)为其他的节点提供服务,有效地减小了服务器的负载。因此,基于P2P网络模型的流媒体服务具有更好的扩展性、鲁棒性和实用性。本文主要研究P2P流媒体点播系统的相关理论与技术。分析了当前比较典型的P2PVoD系统的性能,并在此基础上提出了一种新的基于P2P的视频点播系统模型。它考虑了节点均衡负载对系统整体性能的影响,采用了服务器集中调度与节点分布协调管理相结合的资源定位方式,灵活的候选父节点策略使节点失效后能进行快速的失效恢复。采用了单源单路径与多路径相结合的内容分发策略,系统适用于从任意数据块开始请求。文中对该模型进行了详细的理论分析,对相关算法进行了详尽的描述。最后在局域网中实现了该系统的原型,并进行了测试与分析,测试结果表明原型系统基本达到了设计要求。

【Abstract】 Video-on-demand (VoD) is a user-demand driven based video system. The key point for the large-scale application of VoD system is dependent on how to improve the scalability and reliability in the dynamic environment. The traditional VoD systems are based on C/S (Client/Server) mode, which dispatch a piece of data stream for each request. It cannot ensure the quality of service while the number of users increases with large-scale, especially for popular programme ordering. Therefore, it cannot meet the requirements of a large number of VoD services. To address such problem, the P2P dased distribution technology for streaming media is presented. By utilizing the idle resources(bandwidth, CPU and storage resources, etc) of common nodes(peer nodes) to provide service for other nodes, P2P streaming effectively reduces the load on the server. As a result, the streaming media services based on P2P network model shows better scalability, robustness and practicality.P2PVoD system’s theory and related technology are studied in this paper. The performance of currently existing typical P2PVoD are analyzed, and a new P2P(peer-to-peer)-based VoD system is presented. In the proposed P2PVoD system, the influence of balanced for the nodes on overall performance is taken into consideration, a new resources location mode by combining server scheduling with coordination management for node distribution is adopted, while the flexible strategy for the candidate father nodes ensures a quick failure recovery after the node invalidation. It use the content distribution strategies by combining single-source multi-path and single-path. The system is suitable for request from any data block.The detailed theoretical analysis on the model is applied in this paper, while the related algorithms are introduced in detail. Finally, the prototype of the system is realized in LAN, some experiments are conducted and analysis is employed. The experimental results show that the prototype system can meet the basic design requirements.

  • 【网络出版投稿人】 重庆大学
  • 【网络出版年期】2009年 06期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络