The Online P2P Video-on-Demand (VOD) streaming system has become one of the most popular Internet applications in recent years. Together with millions of users and tens of millions of workload, comes a big challenge for quality of service. This system's time-free accessing strategy might bring about far more severe access failure than that caused by flash crowds in traditional IPTV. Why does access failure happen and how is a user's patience towards it? It is a significant problem in system optimization and user experience improvement. This paper conducts a measurement study of access failure in a largest P2P VOD system named PPlive. We first analyze the statistic characteristics of channels' access failure and its correlation with channels' popularity. Then we propose a method to infer the causal factor of access failure for each user by mining users' viewing behavior and channels' broadcasting performance. Finally we discuss the distribution of user patience facing access failure and find that the initial successful accessing experience will increase user patience. All results can help design a new personalized recommendation balancing user interests and playback quality.
展开▼
