File Transfer Algorithm And Optimization In BitTorrent System

In recent years, more and more network applications have widespread based on the P2P network technology, in which the greatest one is file-sharing application. A lot of systems, such as BitTorrent, Azureus and so on, are taken as the representative to the P2P file-sharing software. The network capacity of BT accounts for the entire network traffic flow ratio is extremely high. So, this thesis conducts the thorough research and optimization to the BT technology, how to improve the efficiency of the BT to transfer file between each other, and enhances the performance of the network services, and reduces the network load has a great practical significance.Firstly, this thesis introduced the characteristics of the P2P network in detail, and then gradually drawn out the P2P file-sharing software—the BitTorrent system. We have a multi-analysis of the BitTorrent principle and the core algorithms, carried on a detailed analysis to the fragment selection algorithm and the neighbor selection algorithm, discovered these algorithms can be for further optimization. At the same time, this thesis conducted an in-depth study on the protocols of BT system such as Bencoding, and the communication protocol between the nodes.Secondly, point out the reasons of BitTorrent to produce large quantities of network traffic and increase of the network burden: when one peer begins to select its neighbors, it just selects the nodes from all the nodes in the network in random, not a binding to choice. Therefore those neighbors possibly are under different ISPs, causes large network traffic when they Exchange data. In order to resolve this inter-ISP network traffic, this thesis proposed a technology solution which based on the principle of select neighbors form Local Area Network, which can guarantee the performance of file-sharing systems, and localize the network traffic. In this scheme, most neighbor nodes (not all) are from the local ISP. So, the BitTorrent not only have the best performance but also can reduce the network traffic significantly.Finally, analyze the shortcoming of the content distribution algorithm, it makes document fragments distribute uneven among the network nodes, which impacts on the efficiency of the system. This thesis proposes an algorithm to control the content distribution of the seeds. The algorithm can effectively avoid the document fragments to be distributed uneven in the network, so that all nodes have the similar number of fragments. Simulation results show that the new algorithm enhances the downloading speed of the node and reduces the average download time, also improves the robustness and efficiency of the whole system.