Due to the properties of self Organization, behavior characteristics affect the reliability and performance in P2P file-sharing system significantly. In order to understand this nature, this paper investigated properties and distribution model of users ' behavior characteristics in MAZE system through statistical methods. And the results indicate that upload and download traffic, upload and download bandwidth of users' can be modeled by lognormal mixture distribution; active time and transmission time of users' can be fitted using the lognormal distribution; while the number of requests and the requested number of users' are modeled by the mixed exponential distribution and Pareto distribution.%由于P2P文件共享系统具有自组织的特点,其行为特性很大程度上影响了系统可靠性和性能.为了解这种性质,利用统计方法研究了Maze系统用户行为特性分布模型及特点.结论表明,用户上传流量和下载流量、上传带宽和下载带宽可以使用对数正态混合分布来描述,活跃时间和传输时间可以使用对数正态分布来拟合,而请求次数和被请求次数可以使用指数分布和Pareto分布的混合分布来表达.
展开▼