Research On Co-Allocation Of Grid Resources

Grid computing is a new technology which has flourished quickly in recent years. As the "Third Generation Internet", Grid is a supercomputer that synthesizes the entire Internet, and the goal of grid is to realize nontrivial resources sharing, task collaboration, eliminating islands of information and resources, and sharing all the computation resources, the memory resources, the data resources, the information resources, the knowledge resources, and the expert resources. At present, the grid technology has been used in fields like commerce, medication, education, military, etc, and grid research has been carried out even by the nations.The Grid's goal is to realize resources'efficiently sharing, therefore, the resource management is the core problem in the grid, including the resources organization, localization and assignment, as well as creating process and other related problems. Because grid is distributed, heterogeneous and dynamic, it is much more difficult to realize resource sharing in grids than in traditional systems. Accordingly, it is more difficult to design grid resource management system. It is often the case in the grid that an application requires multiple resources of different types to be allocated simultaneously. The grid resource co-allocation problem is then produces under such background. Research of resource management system and resource co-allocation has much to do with the development of grid.Firstly, the architecture of information grid based on OGSA has been designed according to our project requirements, and the Layered Agent Coalition Resource Management Model has been present. Secondly, the grid management system, mainly including the technologies like resource description, service interface, service deployment, job description, job submission and so on, is realized by using GT4. Moreover, a algorithm of grid resource co-scheduling named RBP is proposed, in view of the grid characteristic, introducing the resource reservation mechanism to guarantee the success of resources co-allocation, and joining the technology of backfill, simultaneously considering the request proposed by the users about the QoS, such as the deadline, budget when allocates resources for the tasks. Finally, in order to evaluate our algorithm, the GridSim, a grid modeling and emulation toolkit, is chosen to emulate our algorithm. The experiment results show that the algorithm present in this paper is feasible, and it improves the performance of grid.