The Research And Implementation Of Replica Optimization Strategies In Grid

Data grid has the capacity for distributed storage and processing, and it can contain a variety of heterogeneous data. Its goal is to set up an environment of heterogeneous and distribution for the mass data to storage, management, access and transmission, and set up architecture and environment for the related services to effectively realize data and resource share. Data grid provides a kind of mechanism to resolve mass multiform data processing. Replica manageemt is key technology in data grid, and it can greatly reduce the transmission delays, increase the efficiency for the data access and processing. The replica can also lessen load for accessing the original nodes, therefore, it optimize the performance of the entire network. Meanwhile, when some nodes become breakdown, the replica can reduce the probability of data loss, and enhance stability.The basic strategy of the replica is to create multiple copies at different locations. The goals of replica management are to effectively manage each copy in the grid environment, and provide transparent and efficient access for the users. The research hot spots in the replica management include replica catalog, replica creation and replica selection and so on.Firstly, the present situation of data grid and replica management was analyzed and the replica management in data grid was researched. Next the position of replica creating namely the replica deployment strategy was deep studied. Replica was deployed using SAA (Simulated Annealing Algorithm) Based on the replica deployment model and the replica management was optimized. Then grids simulators commonly used at present were analyzed and OptorSim is deeply studied. It is also given. to the overall structure ,component function and the Collaborative model between components of OptorSim. Besides, Most Recently Used(MRU) algorithm is proposed in view of the Least Recently Used(LRU) algorithm's shortcoming.Finally, MRU algorithm and the replica deployment strategy based on SAA are simulated respectively, and they were compared with the LRU algorithm. The result of experiment indicates that MRU is better than LRU in some grid environment, and the strategy SAA possesses commonality By comparing the strategy SAA with LRU,SAA can reduce the response time for requesting file, enhance the performance of the system, and optimize to the replica management.