Resource Clustering Based Decentralized Resource Discovery Scheme in Computing Grid

Abstract

Resource discovery is one of the basic services in grid. In large-scale grid environment, traditional centralized resource discovery approaches limit scalability and result performance bottleneck. Based on the peer-to-peer protocol and the idea of resource clustering, a decentralized resource discovery scheme is introduced to overcome the deficiencies of the centralized resource discovery mechanism. Resources with the same resource type and the similar resource performance get together with the resource clusters that are the minimum units of resource management. Based on the resource clusters, the whole resource space is mapped into two index sub-spaces. A hierarchical and P2P hybrid resource discovery framework is developed to make advantage of the centralized and distributed topologies of the Internet. The algorithm of resource discovery is described. Theoretic analysis and simulation result show that this approach discovers desired resource efficiently.