一种层次结构化P2P网络中的负载均衡方法

A Load Balancing Method in Superlayer of Hierarchical DHT-Based P2P Network

相对于扁平结构化P2P网络,层次结构化P2P网络可利用稳定、高性能的超级节点提高P2P网络在动态环境下的性能.然而,超级节点的负载不均是层次结构化P2P网络面临的基本问题之一.对此,作者提出一种超级节点的负载均衡方法,通过分离超级节点负责的关键字空间和负责的叶子节点空间来为均衡负载提供条件,通过采用"力矩平衡原理"来实现兼顾均衡超级节点负责的叶子节点空间和查询请求负载.实验结果表明:在节点承载容量服从Zipf分布和查找请求服从正态分布或Pareto分布的环境下,负载均衡方法可使超级节点的负载达到较好的均衡,实现了用较少的超级节点承担较大的负载总量.

Compared to flat DHT-based Peer-to-Peer （P2P） networks, hierarchical DHT-based P2P networks can use some stable and powerful peers （called superpeers） to achieve efficient lookup under highly dynamic environments. However, a crucial problem faced by all these net- works is the load imbalance among superpeers. This paper proposes a novel load balancing algorithm, in which each superpeer, besides being responsible for the Chord identifier interval from its predecessor to it, maintains the linear leaf-peer interval within which its leaf peers fall. On this basis, ＂moment balance equation＂ is applied to finding a tradeoff point between balancing linear leaf-peer interval and balancing request loads for a superpeer. Analysis and simulation results show that the method can balance the load among superpeers in proportion to their capacity under Zipf capacities distribution and Gaussian or Pareto requests distribution.