The Study On Resource Management Model In Computational Grid

Computational Grid, with the rapid development of Internet technique, is a new kind of computational mode aimed at a large scale distributed calculation. The resources in computational grid are distributed, autonomic, isomerous and dynamic. The selection of a kind of adapted resource management model to manage grid resources effectively is a key to the implementation of high-powered cooperated computation. This thesis is a study on resource management model in computational grid, and the most typical ones,Hierarchical Model, Abstract Owner Model, and Computational Market/Economy Model,are analyzed deeply here. Hierarchical Model is clear in hierarchy. It settles properly the problems of both site autonomy and isomerous resources and supports resource co-allocation to some extent. The resource agent(Abstract Owner) acts as the media of the users and the resources in Abstract Owner Model. Committing tasks and collecting results follows the mode of order and delivery. The essential of both Hierarchical Model and Abstract Owner Model is integrated in Computational Market/Economy Model, which embodies the computational economy's being applied in grid resource management system. Supply-and-demand principle in market economy is utilized to adjust the benefits between the resource owners and users, guaranteeing mutual benefits in both sides. Every model has its own characteristics. The developing trend of development of resource management model in the future is the mixed model possessing many excellences of other models. In this thesis, a new kind of Center-Management-and-Local-Co-Allocation Resource Management Model has been promoted based on the above comparison. In this Model, by using centralized resource organization and monitor including collecting various resource information, establishing a schedule of resolving user's application problem, and recommending an optimized project of resource allocation and reservation, but not allocating or reserving concretely any resource, Resource Information Service in the grid center can be able to offer dynamically overall resource view at different time from now to the future. The Mobile Agent produced dynamically by grid system is responsible for resource co-allocation and reservation and task coordination, and peer-to-peer idea is introduced, which improves the overloaded network and the decreased communication ability around the grid center server due to centralized resource management and allocation, and liberates the all-level servers and the task entities from the complex works of resource control, task schedule and management etc. The center server can concentrate on collecting variety of resource information and offering referenced solution to the problem. An user's application is viewed as a problem in this Model, accordingly a procedure of performance of the user's application is analogized as a procedure of resolving problem. When an user raises a problem, according to the description of user's problem, the computational grid system establishes a schedule of resolving problem including quality, quantity and schedules of resources to meet the requirement of resolving the problem. This is the primary foundation of establishment and modification of resource allocation. According to the circumstance of current resource available and expectantly usable resource, the computational grid system will find the solution space of the problem, the schedule of resource allocation and reservation. One or a set of optimized solutions are found through comparison. Both the schedule of resolving problem and the optimized solution are inlayed in the Mobile Agent(MA) of this problem produced dynamically, and are sent to the first node in the optimized solution. MA, acting as the full-fledged agent of resolving problem, is responsible for resource co-allocation, negotiation of resource reservation, and organization and coordination of the whole process of resolving problem. MA on the first node selects a current practical optimized solution from a set of theoretically optimized solutions recommended by system through peer-to-peer negotiation with all possible nodes involved in resolving the problem. This is the schedule of resource allocation and reservation restricted by time and space. If the plan can come true smoothly, the problem will be resolved, and the final result...