SubCounter:一种基于语义簇聚的节点子集规模估计方法

SubCounter:A Node Subset Size Estimation Approach Based on Semantic Clustering

为进一步改进性能,很多P2P应用需要系统中各节点子集的规模信息。已有的节点子集规模估计方法主要基于对已有的系统节点规模估计方法的直接应用。本文提出了SubCounter——一种基于语义簇聚的节点子集规模估计方法。SunCounter通过节点间周期性的邻居交换为每个节点维护一个语义簇聚的邻居列表,以保持每个节点与自己所属各子集其他节点的联系。基于这种联系,SubCounter以反熵聚集的方式实现节点子集规模估计。实验结果表明,相比于已有方法,SubCounter在节点所属子集较多的情况下具有更快的收敛速度,并且能够以更小的通讯与存储开销保证同样的估计精度和相似的鲁棒性。

Many P2P applications need the size values of node subsets in the system to enhance performance.The existing subset size estimation approaches are based on applying the size estimation approach directly.This paper proposes SubCounter,a node subset size estimation approach based on semantic clustering.SubCounter maintains a semantic clustering neighbor list for each node by view exchange,so each node can keep contacts with others in the same subset.Based on the semantic clustering,SubCounter realizes the estimation of subset sizes,through anti-entropy aggregation.The experimental results show that compared with the existing approaches,SubCounter converges more quickly when each node belongs to many subsets simultaneously,and ensures the same precision and similar robustness with less communication and storage cost.