in P2P video-on-demand (VoD) streaming system, aggregate download bandwidth of users fluctuates significantly because of peer departure, bandwidth heterogeneity and network delay. The playback rate of videos also fluctuates in order to attain constant image quality. These two kinds of dynamics cause playback interruptions. Prefetching some video content before playback is an effective method to offset the adverse effect of dynamics. In this paper, we choose the number of interruptions in a viewing process as the metric of user perceived QoS (UPQoS). Then we propose a stochastic model to relate the UPQoS with some important system parameters, such as playback rate, downloading speed and prefetched content volume (PCV). It is found that reducing the system dynamics or increasing the PCV or aggregate download bandwidth can improve UPQoS. Besides, the interruption occurs at the beginning of a viewing process with great probability; a few users suffer many interruptions in a viewing process while most of users experience no interruption. Finally, we do a large scale network measurement towards PPLive VoD streaming system and the measurement result is consistent with the conclusion got from this model.