The Study Of The Grid Resource Discovery Method Based On Hierarchical Model

With the increase of high-performance demand, a single high-performance computer cannot solve some of the applications with massive computing requests. In this case, the distributed and heterologious resources need to be connected by high-speed networks in order to share sources, eliminate the "information island" and solve the great-scale applications.Before the emergence of grid technology, the resources in traditional high performance computing systems are distributed in a limited area. It is convenient to search and locate all resources reliably in a very short time. So, in that time, resource discovery has little effect on the application of resources. After the grids having come forth, the resources are distributed widely around the world, the communications delay and the limited bandwidth causes the network unreliable, and thus both the performance and the accessibility of grids are directly determined by the function of resource discovery.The grid resource discovery technology stands as a bridge between resource requester and resource provider. Only with the help of scheme of resource discovery can requesters reach the resources of the grid conveniently. Otherwise, requesters have difficulties in finding proper grid resources, even though there are abundant resources.Firstly, the grid is described briefly in this thesis, such as the characters and structure of grid, the characters of grid resource and the models of resource management. The resource discovery technology used in the MDS of Globus and Vega are introduced and the advantages and disadvantages of central and distributed scheme are presented.Then, this thesis proposes a method of grid resource discovery based on hierarchical model by combining the centralized scheme with distributed one. The resources are divided into three levels, physical network layer, resource information layer, and index information layer. In resource information layer, the virtual organizations are used to manage the grid resources, and one super node is set in each virtual organization in order to store adjacency list and information library. An information library stores information of types and numbers resources in its own virtual organization. The adjacency list is set to manage resources according to their types in virtual organization. Therefore, on one hand, the problem of bottleneck in central scheme is avoided. On the other hand, the disadvantage in distributed...