A Measurement Study Of Quality Of Service In A Large-scale P2P Live Streaming System

In recent years, with its excellent speciality such as scalability and load-balancing,peer-to-peer (P2P) was introduced into a lot of streaming media technique, and became themainstream framework. Meanwhile, with the constant improvement of the networkenvironment, users’ requirement of video service quality has been increasingly demanding.This paper conduct an in-depth analysis of a typical mesh-pull P2P streaming system,namely PPLive. Based on this, we developed a dedicated measurement platform, andfurther conducted a quantitative analysis to comprehensively infer the video streamingquality of remote peers in P2P live streaming.The traditional research method for assessing the quality of service in P2P streamingincludes server logs analysis and network traffic monitoring, which are often severelyrestricted by the software and hardware condition. Moreover, the research of the videostreaming service quality mainly focus on the acquisition and analysis of the originalvideo data. This method is accurate, but can result in a large quantity of network flow.This paper proposes a method of measuring peer’s buffer map to comprehensively inferthe video streaming quality of remote peers in terms of the streaming bit rate, the startuplatency, the playback delay and the video continuity. This paper has completed thefollowing main work: Firstly, we design and present a measurement strategy and a videostreaming quality analysis model; secondly, we make a detailed discussion about designand implementation of the active measurement system; finally, we make an evaluationabout the efficiency of the measurement system, and further conducted a quantitativestatistic analysis of PPLive, and then proposed some advice on the improvement of peerselection and original offset emplacement strategy in mesh-pull P2P live streamingsystem.Through a comprehensive analysis of the measurement result, the accuracy of qualityanalysis model is verified by the experiment and the statistical data also lead to someimportant suggestion for improvement of the P2P live streaming system. Ourmeasurement system is able to operate in a stable manner for long time, and also cancontinuously provide data analysis for the following research, which put forward somevaluable suggestions to the development and deployment of future P2P streaming systems.