A Live Streaming Model Research Based On P2P Network

With the rapid development of internet technology and continued growth of demand for network contents services, streaming media applications in the network services occupy an increasing important position. When using traditional C/S model to build a streaming media services, it will faced some serious difficulties, such as heavy server load, high cost and poor system Extensibility, so C/S model unable to meet large-scale streaming media applications demand. While introducing P2P thinking into streaming media applications can solve these problems effectively.In this dissertation, we firstly introduce the basic theory and development of P2P streaming media applications. then we focuses on tree-based live streaming media model and mesh-based live streaming media model, analyzing the data scheduling defects of a typical mesh model. Based on the above study, we mainly research on the design of a mesh-based live model called FastLive, and achieve some valuable results as follows:1. By analyzing the defects of a typical and traditional pull data scheduling, we propose a push and pull data scheduling, which proved can reduce the delay and redundancy of the data effectively. Further more, we employ a more simple and light algorithm in bandwidth estimation that help to reduce the control overhead of the nodes.2. In order to increase the proliferation of the rare data in overlay, we propose a rare-data disk cache policy. Besides, anther policy called popular-data disk cache also employed, which can further improve the quality of streaming media service.3. The above algorithm and policy are integrated into the FastLive live model, and simulated with GM-Sim simulator. The simulation results show it can significant reduce the data delay and redundant, increase the spread of the rare data and improve the quality of streaming media service. Consummate some of the GM-Sim components among the simulate periods.