Reliability-aware Replica Selection for Data-intensive Applications on Data Grids

Data grids manage and store huge and widely distributed data efficiently to support data-intensive applications. Replication technique used in data grids creates multiple replicas for a dataset. One of the fundamental problems concerning replication is to select the best replica for a dataset to meet the performance requirements of applications. This paper focuses on replica optimization for unreliable data grid environments and proposes a reliability-aware replica selection algorithm. This algorithm considers especially the influence of the reliability of a replica on the completion time and the execution cost of an application. Our experiment results demonstrate that the proposed algorithm achieves the good results in an unreliable data grid environment.