【作者】 唐亮；

【导师】 孙名松；

【作者基本信息】 哈尔滨理工大学 ， 计算机应用技术， 2008， 硕士

【摘要】 如何能在保持系统整体性能的同时,有效地进行资源存储和定位,是P2P点播系统中所需要解决的重要问题。在P2P点播系统的服务器、代理服务器以及客户端中,缓存技术在缓解网络传输压力、平衡系统负载以及减少网络带宽消耗等方面扮演着重要角色,设计和选择适当的缓存策略可以改善系统性能。本文研究重点是在基于有服务器结构对等网络环境的点播系统中,针对客户端缓存技术的研究还比较少的情况,提出了一种新的适合客户端的缓存替换策略。随后给出了该策略的模型,并对该模型进行了详细的阐述,分析了其优缺点。预测双缓存模型具有四个缓存队列,较传统点播系统多了三个缓存队列,其目的在于通过缓存更多有价值的媒体数据来提高缓存的命中率,进而提高系统的性能。然后对预测双缓存替换算法进行了研究,解决了缓存副本分布、缓存副本概率一致性以及阈值的选择等问题。接着又对缓存中数据调度问题进行了深入的讨论,给出了各个不同缓存的数据替换和调度算法,针对缓存调度以及VCR操作引起的数据调度分别给出了解决方案。为了分析对比本文提出的预测双缓存模型与传统的点播系统中的缓存模型优缺点以及证明该模型的可行性和正确性,在分析了传统的仿真器的基础上,利用VC开发出了一个适合预测双缓存模型以及传统缓存模型的仿真器。文中先分析了仿真器的系统结构,对每个模块的实现给出了较为详细的描述,并对所需参数的设置进行了说明。最后使用该仿真器证明了本文提出的预测双缓存模型,能够在点播人数逐渐增多的同时,提高节目内部数据的点播概率,有效缩短VCR操作的时间,提高点播系统的服务质量。更多还原

【Abstract】 How to actualize resource storage and location effectively while the systematic entirety is kept is an important problem that requires to be solved in peer to peer video on demand. In server, proxy server and client of peer to peer video to demand, cache technology plays an important role on aspects of relieving network transmission pressure, balancing overall load, and reducing network bandwidth consumption. Design and selection of caching strategies can improve system performance.This paper focuses on the new cache replacement policy that is fit for client and is proposed towards the fact that few researches on client cache technology have been made and is based on the video on demand with server structure P2P network environment. Later, this paper proposes the model of this strategy, elaborates the model, and analyze its advantages and disadvantages. Double predictive caching model is composed of four cache queries, three more than the traditional video on demand, aiming at increasing hit ratio of cache through caching more valuable streaming data, and then promoting system performance. Then, this paper makes a research on double predictive caching replacement algorithm, which has solved the problems such as cached copy distribution, cached copy probabilistic consistency and selection of threshold. Then, this paper also makes a deep discussion on the problem of data scheduling in cache, puts forward the data substitutions and scheduling algorithms of different caches, and also proposes the solutions for the data scheduling resulted from cache scheduling and VCR operation.In order to analyze and compare the advantages and disadvantages of double predictive caching model proposed in this paper and the model in traditional video on demand, and to prove the feasibility and correctness of this model, this paper develops a simulator which is fit for both double predictive caching model and traditional cache model by using of VC based on the analysis of traditional simulator. This paper analyses the system structure of simulator firstly, gives a detailed description of actualization of each module and also makes an explanation of setting up of required parameter. Finally, the use of the stimulator has proved that the double predictive caching model proposed in this paper is capable of increasing order probability of inner data, effectively shortening time of VCR operation and promoting QoS of video on demand when the number of people who order is increased gradually.更多还原