The Mechanism Research On Similar Interest Cluster Based Grid Resource Discovery

Grid represents an advanced technology and infrastructure, global scale computing resources, storage resources, data resources, information resources, knowledge resources, expert resources and equipment resource sharing can be achieved. With large-scale development to the grid, integrated resource in systems will have the characteristics of the current P2P networks. Unreliable and intermittent resources will account for a considerable proportion of the resources, resource types will be also more varied.Effective organization of resources and select a good resource discovery mechanism, is a key problem in grid resource sharing and collaborative working. After studying, the peer-to-peer character and dynamic variability of grid nodes have similarities with the real world society, each node will have an interest and like. Based on depth analysis of advantages and disadvantages of the current resource discovery mechanism, mining similarity between interest characters of the nodes, to find out the cause and nature of the reasons of nodes network connectivity and resource sharing. A similar character model of the node collaborative learning mechanism is proposed, this mechanism can gain overall similar character information of the whole nodes through limited local knowledge, and it is effective for the polymerization of nodes.Similarity of nodes interests is matched with similar character model, nodes are divided according to interest relations. Connectivity is established above the overlay networks, and a hierarchical Small-World network based on users'interests is constructed. Similar interest cluster based algorithm is proposed, and a certain strategies to measure the probability satisfied inquiries is designed to gain high rates of resource localization in the location at low communication cost.Finally, the simulation is carried through, to evaluate and measure the integrated performance of our resource discovery mechanism, from the search success rate, the search expenses, adding the node bandwidth, search efficiency and scalability, and so on. The learning process and convergence speed are analyzed.