Research On Qea-based Replica Creation Strategy In Data Grid

By creating one or more data replicas and assigning them to different sites, data grid replica management (DGRM) technology can make grid users obtain their requiring data from local or nearby sites, which can not only shorten data response time, reduce user access latency but also save the network bandwidth, balance the server load and improve the data availability.Replica creation is the basis of replica management. Affected by the own characteristics of grid, replica creation in data grid has been proven to be a NP problem. Quantum Evolutionary Algorithm (QEA) combines the advantages of both the quantum computing and evolutionary computing, and it has better search capability and convergence speeds compared with the traditional evolutionary algorithms. In recent years, QEA has been successfully used in many fields. However, it has not been seen that QEA is employed in DGRM.In view of the above analysis, this thesis mainly researches on QEA and its applications in data grid replica creation, the major contributions are as follows:(1) The research status of DGRM and QEA are reviewed, respectively. Then data gird and DGRM technologies are discussed in details. And the applications of computational intelligence algorithms in DGRM are also anlysized. Meanwhile, an abstract model is built to describe DGRM. Based on the model, the key issues of DGRM are depicted clearly as well as the working flow and cooperative relations among them.(2) The basic principles of QEA are depicted in details. By defining the conception of Angle-distance between qubits, a new Variable Angle-distance Rotation (VAR) strategy is designed to dynamically adjust the rotation angle. Base on VAR strategy, a novel quantum evolutionary algorithm is proposed called as QEA-VAR. The experiment results on the basis of the constraint0-1knapsack problem show that QEA-VAR has faster convergence and better profits compared with other algorithms.(3) The replica creation models of single data and multi data are defined as foundations, respectively. After that, a new data grid replica creation strategy based on QEA-VAR is designed. Meanwhile, some simulation experiments are carried out by expanding OptorSim to demonstrate the performance of the designed strategy. The results show that it can effectively reduce the mean job execution times and consumption of network bandwidth.