P2P网络节点自组织聚类算法

Node clustering algorithm in self-organizing mode for P2P network

提出了一种P2P网络节点自组织聚类算法,该算法具有分布式、自组织的特点,每个节点分别使用自身掌握的网络局部视图,通过邻近节点查找机制独立地完成聚类操作,为P2P网络构造算法提供支持,具有良好的可扩展性和鲁棒性。为提高邻近节点查找效率,节点根据小世界模型在聚类区域外采用半径指数递增且互不重叠的多重环结构组织远距离节点,增加捷径连接,减小节点间平均距离。在邻近节点查找过程中,使用分布式的网络坐标机制预测网络距离,缩小目标节点范围,然后通过直接测量找出最邻近节点,这种策略在保证准确度的同时有效地降低了系统开销。

A node clustering algorithm is proposed in self-organizing mode for P2P network.The algorithm is fully distributed and selforganized.By analyzing the local view of network and employing a nearest neighbor searching scheme,each node constructs independently a cluster of which its own is the center.As a result,the algorithm has high scalability and reliability.In order to achieve efficient clustering,each node maintains a data structure which puts the set of other peers,who are not covered by the cluster,into concentric and non-overlapping rings whose radiuses are exponential increased based on the small world model.In the searching process,the algorithm designs an approach which performs direct RTT measurements based on the result of distance prediction using network coordinate,and both efficiency and accuracy are achieved.