Data Grid Optimization using Replica Replacement

The Data Grid provides computation and storage facilities for intensive applications to carry out their computational jobs. The computational jobs in grid is carried out by grid scheduling, file access pattern and replica replacement and hence the performance of the grid is depending on them. The replication in the grid plays a major role in data grid. As the storage capacity is limited, replica replacement is required for providing space for new replication. The replica replacement should increase the performance of the grid. The replica replacement algorithms include LFU, LRU, Binomial, Zip f algorithms are using the uni-dimensional parameters to predict the possibility of using the same replica in the near future. But, it is a need for replica replacement algorithm to use the multidimensional parameters to predict the future use of the replica. The proposed work is using multi-dimension parameters to predict the future use of the replica. The proposed replica replacement algorithm using normalization and weight values to predict the future usage of replica. The proposed algorithm is simulated in the OptorSim simulator. The Hit Rate, Mean Job Execution Time, Effective Network Usage, and Computation Element Usage are used for evaluating the performance of the algorithm. The simulation result is showed that the proposed algorithm yielded better performance in comparison with other algorithms.