Analysis and design of peer-to-peer video on-demand streaming

Abstract

The problem of video on-demand (VoD) streaming via peer-to-peer (P2P) overlay networks is still challenging. A novel resilient P2P VoD streaming system is proposed in this thesis based on a multi-video-block transport framework with user interactivity supports. The proposed system, namely resilient VoD (rVoD), consists of a distributed encoded video content caching scheme and a transport protocol based on rateless codes. Notions of production and saturation time, naturally arising in this context, are introduced to characterize the dynamics and performance of the P2P streaming system. Quantified in terms of production, the throughput of the presented framework is shown analytically to increase with time in a manner no worse than an exponential function prior to the saturation time. The notions of production and saturation time may also serve as fundamental concepts for the study of quality of service (QoS) for the proposed system. A necessary condition of saturation time and the relationship between production and server stress are derived for an optimal P2P streaming system. We then study the definition of QoS metrics and their tradeoff. A constraint equation of QoS metrics tradeoff is derived, and the numeric results are demonstrated. A simulation system is developed to measure the performance of rVoD in terms of production, saturation time, several QoS metrics, server stress, latency, and overhead. Results of the simulations demonstrate the advantages of the proposed system compared to previous works and confirm our theoretical analysis.