| This thesis mainly discusses the implementation and the simulation of a P2P network file sharing system, which is based on a P2P network simulation platform on PDNS. On this foundation, this thesis designs and implements a few strategies for some key algorithm module, processes parallel simulation to compare and analysis performance of these strategies. The purpose is to learn about how the performance of this system is influenced when some factors are changed. We mainly focus on two key algorithm modules of the system. The first one is piece selection policy and the second one is Tracker mechanism.This thesis implements three strategies for piece selection. They are Rarest First policy, Random First policy and Region First policy respectively. Except the traditional random peerlist selection policy, this thesis also implements a new strategy for Tracker mechanism, which is called"Smart Tracker"mechanism. The feature of this mechanism is to help peer find more peers whose bandwidth is the same as it and exchange data to each other. It improves the speed of download and the justice of system in a certain extent.In the process of testing the performance of each strategy, many metrics and parameters have been used to analyse result of simulation: average download time, distribution of download time, normalization of seed, utilization of seed, Number of interesting connections, piece inter-arrival time, etc. These parameters help greatly to analyse the differences between different strategies. Meanwhile, large number of charts help researchers more intuitive to find out the reason of casuing these differences.This thesis implements a P2P network file sharing system and analyses the performance of the system roundly when simulating in different strategies. Given the statistic result of the simulation, partial improvements in the algorithm are made. The methods and the results presented in this thesis provide important reference to the P2P network file sharing research and application. |
PDF Full Text Request