一种基于K-means聚类分组的P2P超结点模型  

A SUPER-PEER P2P MODEL BASED ON K-MEANS CLUSTERING

作　　者：江华[1] 王翰虎[1] 陈梅[1] 

机构地区：[1]贵州大学计算机科学与工程学院,贵州贵阳550025

出　　处：《计算机应用与软件》2005年第10期67-68,86,共3页Computer Applications and Software

基　　金：教育部春晖计划项目

摘　　要：本文面向网格计算、协同工作等特定P2P应用,提出了一种基于超节点的P2P混合网络模型。该模型的主要特点是采用K-means聚类算法对网络中的节点进行聚类,从而得到分组虚拟子网,由各子网内的超节点负责临近节点的加入与注销。该模型既克服P2P集中模型伸缩性不强,分散性P2P模型发送消息量大等缺点,又可以灵活地组织对等节点进行网格计算或组内协同工作等分布式应用。文章最后使用仿真实验表明K-means算法在网络节点聚类中的有效性。The paper described a hybrid P2P model with super-peer that is oriented on the special P2P applications, such as grid computing and cooperation work. The K-means algorism is utilized to cluster the nodes in the network in this model, and every cluster is a virtual sub network. The main function of super-peers is responsible for the registration, logout of the new peers, maintaining the index and network distance map. The model does not only improve the flexibility of the distributed computing platform, but also reduces the request messages. The model is conducive to distributed applications. The result of simulated test is described in the end of the paper that showed the efficiency of using K-means algorism to clustering network hosts.

关 键 词：混合P2P超节点 K-MEANS聚类 分布式应用 K-MEANS聚类算法 网络模型 P2P 分组 超结点 K-MEANS算法 网格计算 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP18[自动化与计算机技术—计算机科学与技术]