Abstract:  In order to further improve the accuracy of video on demand model for system behavior prediction, a method considering fault node based Model with non cooperative peer to peer assisted for video on demand is proposed. Firstly, the finite population representation model was constructed, which used the idle queue、the receiving queue and the failed queue as the components; Secondly, the peer to peer (p2p) network was used to describe the network behavior of video on demand system with three queue, and the numerical model of the system was also built, which gives the system state probability equation coefficient calculation process and system performance calculation formula; Finally, the experimental results show that comparing algorithms on the two indicators of uplink traffic utilization and service utilization is that the proposed method is superior to all selected, while the experimental part of the system is given a node failure performance of experiments.

Key words: Fault node, peer to peer network, poisson distribution, streaming media, video on demand

摘要： 为提高视频点播模型对系统行为预测的准确性，提出一种考虑故障对等节点的非合作对等网络视频点播建模方法。首先，构建具有泊松分布过程的idle队列、接收队列和故障队列的视频点播的有限种群排队系统表示模型；其次，利用对等网络(p2p)对视频点播三队列系统的对等辅助流媒体的网络行为进行描述，构建系统数值模型，并给出系统状态概率方程系数计算过程和系统性能计算公式；最后，通过实验对模型预测精度进行验证，仿真结果表明在上行链接利用率和服务流量利用率两项指标上，所提算法均要优于选取的对比算法，同时实验部分给出了节点故障对系统性能的影响实验。

关键词: 故障节点, 对等网络, 泊松分布, 流媒体, 视频点播