基于坐标的网络节点聚类在Internet中的实验研究

在分布式网络应用中,节点聚类是构建高效网络体系结构的有效办法. 网络坐标能够反映节点在Internet中的位置,利用网络坐标可以对网络节点之间的距离进行预测. 在测量各个节点的网络坐标的基础上,利用Triangulated heuristic方法对节点间的网络距离进行预测,提出了网络距离作为参数对网络节点进行分布式聚类的算法,并在PlanetLab分布式实验床遍布全球的156个节点上进行了实际的测量和分析. 分析结果表明,算法能够在网络规模较大的情况下对网络节点进行分布式聚类,算法有较高的可靠性和可扩展性.

Experimental study on network coordinate-based node clustering

In the application of large-scale distributed network,node clustering is a useful way to construct an effective network infrastructure.Network coordinate can reflect node's position in Internet and can be used to predict the network distance.After getting every node's network coordinate,the authors use Triangulated heuristic to predict the network distance between nodes.The authors present a distributed clustering algorithm using network distance,and having done the experiments on 156 nodes of PlanetLab testbed all over the world.The experimental results show that this algorithm can work well on large-scale node clustering in distributed way,and it has high reliability and scalability.