Study On Modeling Analysis And Optimization For Offloading Servers In Large-scale P2p-vod Systems

Peer-to-Peer Video-on-Demand systems, as new Internet applications, are broadly popular. When compared with P2P live streaming systems, peers in P2P-VoD systems, however, are interested in different videos. Though some may be watching the same video, they even playback the different segments of the video. As logical consequence of peers'asynchrony diluting the ability to assist each other, they consume more bandwidth resource of server and lead to a bottleneck at server.The bottleneck at server poses severe challenge to improving the recent P2P-VoD systems. Therefore, we endeavor to find the breakpoint by thoroughly understanding the rationale of systems so as to alleviate heavy server load. Firstly, a probability-based server load calculating model is proposed to understand the relations between server load and critical parameters, such as peers'upload capacity and cache capacity. Then, a peers'contribution model is developed to analyze contribution capacity among peers and the relations between contribution capacity and critical parameters. Both models are verified by simulations in the corresponding scenarios.We then attempt to respectively employ helper network and network coding to alleviate the server bottleneck. On the one hand, helper network is introduced into the recent P2P-VoD systems. To understand the performance of P2P-VoD systems with helper network, we develop analytical model and derive the optimal download amount of helpers under the circumstance that all segments are uniformly downloaded. Simulations validate our analytical results. We conclude that helper network can efficiently alleviate server load by serving as a bandwidth amplifier.On the other hand, we employ network coding to improve the recent P2P-VoD systems. The aforementioned probability-based server load calculating model is utilized to calculate the server load and compare with the recent P2P-VoD systems. Simulations validate the calculating model and verify the benefits brought by network coding. We conclude that network coding can drastically increase the diversity of peers'caches, especially in the case that peer's cache capacity is limited, therefore network coding can remarkably decrease the server bandwidth cost.