The Research And Implementation Of A Network Coding Based P2P Content Distribution System RlncBT

P2P mode is more and more applied to the area such as content distribution, instant messaging, distributed computing, teamwork and so on, because of its decentralization, scalability, robustness and load balancing. P2P technique has also brought new development and applications in these fields. Among them, the content distribution is always the driving force of network techniques and one of the most typical P2P applications.At present, BitTorrent is a very popular P2P content distribution system. However, it needs improvement in the following aspects: one of its serious problems is that BitTorrent has strong dependency on the seeds. If on-line time of seeds is relatively short, the successful downloading ratio would be very low and it would cost a longer time to complete the downloading. The utilization ratio of the network bandwidths is not high, because the connection between any two peers is often idle. With the network scale expanded and the number of download participating peers increased, the rarest first piece selection algorithm has got no obvious vantage.As a new technique in the communication field, network coding has got extensive attention. How to get high performance of P2P content distribution by making the best use of network coding is the main research task of this thesis. Aiming at the shortcomings of the present BitTorrent, this paper first researches on the core ideas and the key algorithms of BitTorrent and the basic theory and feasibility of network coding, especially in the efficient application of random linear network coding. Then, based on the modular design principle, this thesis designs and implements a network coding based P2P content distribution system rlncBT and lays the emphasis on discriptions of the following modules in rlncBT client: the main control module, the Tracker sever interaction module, the file processing module, the random linear network coding and decoding module and the peer interaction module. Test results obtained by GPS simulator show that if the seeds leave the system early and peers arrive and depart frequently, rlncBT has a faster downloading rate, higher downloading success ratio and better robustness than BitTorrent.