Research On Grid Resource Discovery Mechanism

Grid is a newly developed decentralized computing based on Internet. Its basic goal is to coordinate and share resource, but attain this goal on the premise that resource discovery. Resource discovery finds the resource and service to meet the demand of application. Grid is not only facing traditional computing resource, but more is to face the interaction between users. However, traditional resource discovery mechanism which organizes resource based on the feature of resource fails to take interaction into account, so it has not evident effect on improving efficiency of resource discovery.Based on the research of grid resource discovery mechanism of hierarchy and P2P resource discovery mechanism, this paper designs a resource organization structure which using virtual organization as its basic units to overcome the disadvantages of less reliable and scalable of centralized resource discovery mechanism and inefficient of distributed resource discovery mechanism. This structure has both advantages of centralized structure and distributed structure. It includes index service layer and network layer logically, the top layer is index service layer and the lower layer is network layer. Based on the structure, this paper proposes a resource discovery model based on user interaction feature to overcome the disadvantages of traditional resource discovery mechanism which organizes resource based on the feature of resource. It finds the user interaction feature from the history record. The resource feature relates to corresponding frequent interaction nodes will join the same virtual organization in the model. On the base of analyze users' interaction, the model builds the virtual organization which provides the users' favorite resource, so the requests are satisfied mostly in the virtual organization and the requests which are not frequent are satisfied in other virtual organizations, therefore, it will reduce search space, improve the efficiency of resource discovery and reduce the cost of communication in the backbone network. Finally, the effectiveness and high efficiency of the model were validated through the experiments.