Measurement and Modeling for QoS of VOD System

It is still an open issue to evaluate the quality of service (QoS) of a P2P VOD streaming system. In this paper, we use fluency rate (FR) as the metric to reveal the system ability of playing videos smoothly. From the perspective of user experience, we use average playback interruption times to estimate the customer satisfaction of a VOD streaming system. Further, we analyze the distribution of playback interruption positions within a video and find that the playback interruption probability at the beginning of videos is much higher than that at other positions. After analyzing the coding rate of thousands of videos, we use Brownian motion with drift to model videos' playback process and peers' download process. Then we present a mathematical model to compute the playback interruption probability of videos during the whole viewing process. This model matches the network measurement result and is beneficial to design new pre-fetch strategy, video slicing strategy and so on.