Research Of Replica Selection Strategies In Data Grid

In Data Grid, multiple replicas of data files are created to improve data reliability and reduce data access time and bandwidth consumption. In such an environment, the data files user applications needed most possibly locate in distributed grid nodes. So it's critical for users to select the best replicas, based on their performance and data access features, to access the needed data files fastly. In this paper, the author proposed a replica selection strategy based on Ant Algorithm, which offered a new novel method for replica selection in Data Grid.Firstly, the paper put forward the original replica selection strategy based on the ant algorithm. The correcting feedback, cooperation, inherent parallelism and cryptic dynamic of Ant Algorithm have made itself suitable for the distributed system and grid environment perfectly, whose architecture and components will alter dynamically. The author designed the replica selection strategy after making a deep research into the ant algorithm and investigating the key factors that influence the performance of replica selection greatly. And then analyzed and implemented this strategy.Secondly, OptorSim, a popular data grid simulation environment, was researched and analyzed to a deep extend. To make an analysis of the proposed strategy under OptorSim, the author modified and extended the replica module, the replica optimizer module, the replica transfer module, the statistics module, the GUI module and other main modules of OptorSim. The ant algorithm based replica selection strategy was added into OptorSim and the extended GUI module displayed the simulation results successfully.Finally, the advantages of the new ant algorithm are investigated in OptorSim after it has been expanded. The author designed three scenarios to compare the new stratety with another replica selection algorithm. The evaluation demonstrates that this new ant algorithm can reduce data access latency, decrease bandwidth consumption and distribute storage site load.The last part summarized all the research work and gave some prospect to the further work.