基于混合聚类的覆盖网络组播服务节点选择模型

A Selection Model for Multicast Service Nodes Based on Hybrid Clustering

通过分析当前覆盖网络特点,以分层覆盖网络为基础构建特定源组播树,实现对组播服务节点MSNs的管理,并提出一种基于K-Medoids和遗传算法的模型,用于网络中MSNs的选择.结果表明,该模型有效克服了传统K-Medoids算法模型对初始中心选值敏感的问题和早熟收敛现象,使其针对覆盖网络组播服务节点的选择性能明显优于K-Medoids选择模型,平均收敛速度也提高近30%.

On the basis of analyzing overlay network, a source-specified multicast tree with hierarchy structure was constructed to manage the multicast service nodes based on NICE network. Based on the algorithms of K-Medoids and genetics, a hybrid K-Medoids genetic model （HKGM） was proposed to choose multicast service nodes （MSNs） from network nodes. The results show that HKGM not only avoids converging to local minimum value, but also is robust to initialization. The performances of selecting MSNs, HKGM is superior to K-Medoids, and the average convergent speed increases about 30 %.