| Peer-to-Peer (P2P) technology has achieved a wide range of applications. As one of the P2P file distribution system, BitTorrent has become the hotspot of P2P study. This thesis aimed at decreasing file download time in order to improve BitTorrent distribution performance. The main research of this thesis is as follows:Firstly, an optimized node selection strategy was proposed to overcome the drawbacks of random node selection in BitTorrent systems: Tracker server classified online nodes useful information, which included the upload, download, and node state. When there was node requesting neighbor list, tracker selected the neighbor nodes by the request node's period partially from the statistical results, partially from the random results. It was worth noting the seed nodes, which will work on the optimized strategy introduced later, if it was contained in the returned neighbor nodes.Secondly, this thesis pointed out that there existed unoccupied bandwidth, which was download bandwidth at the node's initial stage, while upload bandwidth at the last stage of nodes. In response to this, an optimized content distribution strategy was proposed: When node was at the initial or last stage, it will adaptively increase its unoccupied bandwidth in order to effectively integrate the idle resources and improve the system performance.Thirdly, according to seed character, this thesis proposed that the seed node can regulate the piece distribution. To avoid the negative effects of rare piece, the seed node's operating mechanism was optimized: seed node will stat. pieces amount which were owned by neighbor nodes, then calculate the rare pieces, so it can balance the amount of rare pieces by giving priority to the nodes which requests the rare pieces.Based on distributed simulation platform, this thesis designed and implemented these optimized strategies. Simulation results demonstrated that optimal strategies achieved better distributed efficiency while retaining the architecture of BitTorrent. |
PDF Full Text Request