Using service taxonomy to facilitate efficient decentralized grid service discovery

Abstract

In this paper, an efficient decentralized Grid service discovery approach is proposed, which utilizes service taxonomy to address both scalability and efficiency. Grid information nodes are organized into different community overlays according to the types of the services registered, and a DHT (Distributed Hash Table) P2P based upper layer network is constructed to facilitate the organization of community overlays and provide efficient navigation between different communities. As a result, service discovery will be limited into related communities, which improves efficiency. A simple and lightweight greedy search based service location (GSBSL) method is also introduced to identify service providers with high QoS efficiently within the same community. Simulation results show that, the efficiency is improved compared with existing decentralized solutions, and the overhead is acceptable and controllable. © Springer-Verlag 2004.