Agent Based Replica Placement in a Data Grid Environement

In a data grid, large quantities of data files are produced and data replication is applied to reduce data access time. Determining when and where to replicate data in order to meet performance goals in grid systems with many users and files, dynamic network and resource characteristics and changing user behavior is difficult. Therefore efficiency and fast access to replicated data are influenced by the location of the resource holding the replica. In this paper, we present an agent based replica placement algorithm to determine the candidate site for the placement of replica. An agent is deployed at each site holding the master copies of the shared data files. To create a replica, each agent prioritizes the resources in the grid based on the resource configuration, bandwidth in the network and the demand for the replica at their sites and then creates a replica at suitable resource locations. We have carried out the simulation using GridSim Toolkit-4.0 for EU Data Grid Testbed1. The simulation results show that the aggregated data transfer time and the execution time for jobs at various resources is less for agent based replica placement.