结构化P2P系统中基于网络定位的负载均衡算法*

结构化P2P系统中，各对等节点处理能力的差异以及关键字通常与一定的语义相关，导致系统中节点的负载不均衡。算法针对基于DHT的大规模计算网络中，计算任务在节点间分布不均衡的问题，提出了一种高效的基于网络定位的负载均衡算法：当某个节点的负载较小时，它将以自己为中心，与物理位置相近的节点构成一个星型结构区域，然后在这个物理位置相近的区域进行负载转移。该算法具有扩展性好、效率高、维护简单的特点。仿真实验表明本算法可以达到理想的负载均衡效果，并使负载转移开销减少了40%以上。

Load balancing algorithm based on network positioning in structured P2P systems

In structured P2P systems,the heterogeneity of node capacity and semantic relativity of keys could cause load imbalance among nodes.Aimed at the problem of tasks distributed unbalancedly among nodes on large-scale DHT networks,this paper presented an efficient algorithm based on network positioning.While the load of a node was light,the node,as a center,would construct a star-like structure area with other nodes physically close to it.And then,load could be transferred in that physically close area.This algorithm is scalable,efficient and simple.Simulation experiments show that the algorithm can achieve a good load balance and the load movement cost reduction rate is above 40%.