针对分布式发布/订阅系统的特点,结合结构化P2P网络负载均衡算法的优点,实现了一种基于结构化P2P网络的语义发布/订阅系统的负载均衡算法.采用本体支持语义,定义本体类权重、节点权重、虚拟订阅等概念,将本体类权重树映射到代理权重网络.由代理节点管理虚拟订阅,本体路由表维护映射信息.事件和订阅根据其本体信息被分配到不同的代理节点上.实现了对结构化P2P网络的基于语义的划分,达到分布式计算和匹配的目的.在此基础上,提出一种动态负载均衡算法:动态调整本体类权重及代理节点的权重,在代理节点负载过载时,进行虚拟订阅的交换或迁移,使得系统获得高效的负载均衡.实验结果表明,负载均衡算法显著地改善了结构化P2P代理网络的可用容量,具有良好的可扩展性,在网络资源消耗方面取得了良好的平衡效果.
A dynamic load balancing algorithm of semantic publish/subscribe system over structured P2P networks was proposed by combining characteristics of the distributed publish/subscribe system with advantages of current structured P2P (peer-to-peer) load balancing algorithms. Ontology provided semantic support. Concepts of ontology class weight, broker weight and virtual subscription were defined. Weighted ontology class tree was mapped to the weighted broker network. Brokers managed virtual subscriptions, and ontology routing table maintained the mapping information. Events and subscriptions in accordance with their ontology information were assigned to different brokers. Then the system achieved semantic-based division of the structured P2P network and large-scale distributed computing and matching capabilities. Weights of ontology classes and brokers can be dynamically adjusted, and brokers can exchange or transfer the virtual subscriptions when overloaded. Thereby the system obtained efficient load balancing. Experimental results show that the load balancing algorithms significantly increase the capacity of the broker network and achieve good scalability and good tradeoff of network resource consumption.