针对移动网格的特点提出基于超级节点的覆盖网络模型,对节点聚集问题进行形式化描述。提出的聚集算法以分布式网络坐标系统为基础,由查找聚集、建立新聚集和超级节点连接三部分组成。节点在加入移动网格时首先要寻找最近距离的聚集加入,在网络规模较小时采用简单的禁忌搜索算法来寻找归属聚集,而在网络规模较大时采用改进的粒子群算法寻找归属聚集。在加入聚集时考虑节点间距离和节点容量参数,从而能够满足应用相关的延时要求并实现良好的负载平衡。算法体现了分布式、自适应、轻量级的特点。仿真实验分析了不同参数对算法性能的影响。
Super-node based overlay network model is proposed according to features of mobile grid,and node clustering problem is depicted formally.Clustering algorithm,which is based on distributed network coordinate system,consists of finding cluster,creating new cluster and connecting super-nodes.Node firstly finds nearest cluster when it tries to join mobile grid system.Vicinage searching algorithm is adopted if network size is small,while particle swarm algorithm is adopted contrarily.In order to meet requirements of delay and load balance,distance and capacity parameter are considered when new node joins one cluster.The algorithm proposed is characterized with distributed,self-adaptation and lightweight.Affection of different parameters to algorithm performance is analyzed by simulation experiment.