Dynamic Algorithms Replication Using Grid Computing

Abstract

Data grid is one of the most popular grid computing implementations concerning data management system and data replication technologies. The aim of data replication is to increase availability, fault tolerance, load balancing and scalability while reducing bandwidth consumption, and job execution time. In this paper we investigate Data grid as a solution for the internal network traffic bottleneck at the University of Khartoum due to high numbers of users by replicating the internal systems of the University closer to the users. OptorSim Grid simulator was used to study the behavior of the various dynamic replication algorithms and to evaluate their performance metrics. The results showed that the predictive algorithms are the most effective at optimizing computing and storage resources and they offer the best effective network usage. Other metrics such as the thread numbers at each computing element and the bandwidth range to particular site in comparison to the number of executed jobs were also evaluated. It was found that the thread number was inversely related to the mean job time while the link bandwidth of a particular site was proportionally related to the number of jobs been executed on that site.