A dynamic replication mechanism for data grid federation systems, based on replica placement cost

Abstract

Data replication plays a major role in making data more available in the distributed computing systems, such as data grid federation systems. Also, it is typical of a data replication mechanism to place file replicas in storage locations that offer lower transfer time for data items, without regards to the distance between host and the requesting client. Although file transfer time is a measure of file size and bandwidth usage, which drastically affect storage usage as well as jobs completion time, it requires additional parameters in order to make replica placement more cost effective. In this write-up, a replica placement mechanism is proposed that computes optimum replica placement cost from file value, transfer time and site distance for efficient data replication in a federated data grid environment. The proposed mechanism was evaluated in OptorSim simulator and has effectively improves jobs execution time, bandwidth usage and storage usage within a federated data grid environment, compared to the existing mechanisms.