Design And Simulation Of Peer-to-Peer Live Streaming System

Live media streaming, the most typical multimedia application in Internet, attracts widely interest and study. Peer-to-peer technology, which shows great advantages on lower deployment cost and higher scalability, now became the most popular solution in live streaming.This thesis first discussed the shortcoming of current P2P live streaming systems. We then proposed an unstructured P2P live streaming media system, which based on asynchronous data transfer and centered directory service. Towards peers'heterogeneous in calculation ability and access-layer bandwidth, we adopt one kind of round-robin request strategy with layered video coding, which enables peers adaptive adjustments on transmission rate,to achieve the optimal balance on playback continuity and video quality. We design a traffic localization strategy to reducing network traffic across different ISPs. We also proposed a contribution-based incentive mechanism which ensures fair data sharing among peers.We implement the protocol in the Parallel/Distributed Network Simulator PDNS and discuss the protocol design and implementation details, including the partner management, data transmission scheduling, traffic localization algorithms. Finally, large scale packet-level simulation illustrates that the system could effectively improve system scalability, strengthen the playback quality, decrease server bandwidth stress and reduce network traffic across different ISPs. We also evaluate the performance under different protocol parameters. The simulation results are beneficial to the further study on P2P live media streaming.