摘要： 自适应层次化聚类算法将单层拓扑映射为一个层次化结构，改善了应用层多播算法的可扩展性。但它也引入了较高的根节点度数，无法有效控制多播树的形态，不适合用于多媒体等内容的多播通信。通过“剪枝重置”和“构建顶层拓扑”操作，可以弥补自适应层次化聚类算法的缺陷。仿真试验表明，前者可以减少树的层数和同一个聚类中的节点密度；后者可以减轻多播树根节点的负担。相对于采用传统路由算法得到的单层拓扑多播树，该文得到的多播树在开销、时延和度数方面都有显著的优化。

关键词: 多播, 应用层多播, 聚类, 自适应聚类, 自适应层次化聚类

Abstract: Adaptive hierarchical clustering algorithm (AHCA) maps a flat topology to a hierarchical one while it almost always introduces a bad degree metric and is not suitable for data distribution. Prune-relocate operation and top topologies operation are proposed to make up AHCA’s shortcoming. Numerical simulation shows that prune-relocate operation reduces the total levels and the average number of hosts in a top cluster. Top topologies operation reduces the load of root. The multicast trees by AHCA and the two operations optimize the overall performance of single-level topology protocol.

Key words: Multicast, Application layer multicast, Clustering, Adaptive clustering, Adaptive hierarchical clustering