Design and analysis of a load balancing strategy in Data Grids

Developing Data Grids has increasingly become a major concern to make Grids attractive for a wide range of data-intensive applications. Storage subsystems are most likely to be a performance bottleneck in Data Grids and, therefore, the focus of this paper is to design and evaluate a data-aware load-balancing strategy to improve the global usage of storage resources in Data Grids. We build a model to estimate the response time of job running at a local site or remote site. In light of this model, we can calculate slowdowns imposed on jobs in a Data Grid environment. Next, we propose a load-balancing strategy that aims to balance load of a Data Grid in such a judicious way that computation and storage resources in each site are simultaneously well utilized. We conduct experiments using a simulated Data Grid to analyze the performance of the proposed strategy. Experimental results confirm that our load-balancing strategy can achieve high performance for data-intensive jobs in Data Grid environments.