Research And Optimization On BitTorrent Protocol For Video Application

BitTorrent protocol is a file sharing protocol based on P2P (Peer to Peer), with "the more users the download faster" feature. It reflects the superiority of the system especially for large files. In recent years, Video media streams are a class file that the most amounts of Internet users occupied. With the further increase consumer bandwidth, more and more people want to get as multimedia applications such as video on demand. The traditional C/S architecture, when a sharp increase in the number of users, prone to overload the server, using the server cluster or distributed system cost increases. BitTorrent can better reduce server load and improve the scalability and robustness of the system. But BitTorrent designed for ordinary share large files, select the file fragments are disorderly conduct, and can not effectively support on-demand and live video for large files. In this thesis, video-on-demand needs of large files, fragment selection algorithm to optimize the design of appropriate BitTorrent protocol, developed on the basis of demand and live to achieve a prototype video files.The main works of this thesis include the following aspects:First, this thesis analyzes the working principle of BitTorrent, BitTorrent protocol structure and encoding format, and it has learned the process of file transfer system. On this basis, this thesis studies the salient features of BitTorrent:peer publication, assembly line, piece selection algorithm and blocking algorithm. On the BitTorrent network mathematical model to analyze the impact of BitTorrent network performance elements:network bandwidth, fragment selection algorithm and seed number.Second, this thesis learned the special needs analysis of streaming media technology and users to download video files, streaming media services contrast with the large BitTorrent file distribution system is different:BitTorrent protocol fragment selection algorithm is designed for the common large files, in order to speed of download, it is designed to download documents into disordered. But the video files require rapid and orderly download support. For streaming media services, it proposed the ideas and programs for BitTorrent protocol modification.Again, this thesis is proposed to optimize the design for the BitTorrent fragment selection algorithm. On the one hand, adding speed algorithm to calculate the fastest network node; on the other hand, for the agreement to be redesigned, changing the original algorithm disorderly break download, using a new variable to achieve an orderly download window. Build data models for old algorithm and new algorithm, and analysis algorithms and protocol. The results proved that the new scheme can achieve better demand and live video files.Finally, the thesis set up the experimental environment, and develops appropriate prototype system, and tests the algorithms and protocols and verifies the actual results. It has collected and analysis the experimental data. The results show us that the new algorithm can reduce latency video-on-demand, and improve fragment arrival rate. It can reduce the breakpoint phenomenon playback, and more suitable for video-on-demand and live.