Job-attention Replica Replacement Strategy

Grid computing is increasingly being viewed as the next phase of distributed computing. In this paper we propose a new replica replacement strategy, called JARRS (shorts for job-attention replica replacement strategy), which applies for replica management in data grids. A novel interpretation of locality appropriate for grid is presented. It is the grid locality that lays foundation of replica replacement strategy. A new approach, file forward-looking, is suggested. JARRS uses file forward-looking and references to recently accessed files to make decision. Experiments show that JARRS has a better efficiency and robust property for different file access patterns than others.