Research Of Grid Resource Selection Algorithm Based On Cactus Application

In recent years, the development of Grid is very fast. While Grids can in principle be used as a computing platform for all kinds of application, in practice the problems of selecting the best resource set in Grid and then organizing resources to meet application requirements are difficult. Grid resource selection's purpose is to select suitable grid resources as the computing platform for grid applications.Condor provides a general resource selection mechanism. A matchmaker is used to match user requests with appropriate resources. When multiple resources satisfy a request, a ranking mechanism sorts available resources and selects the best match. Because the matchmaker was designed for selecting a single machine on which to run a job, however, it cannot be easily applied when a job requires multiple resources.In 2002, University of Chicago presents a resource selection algorithm, which extends the Condor matchmaking framework to support both single resource and multiple-resource selection. But because the algorithm selects the most powerful node in Grid at first, it is very easy to fall into partial optimization. When there exist some other resource sets, which doesn't include the most powerful node but have higher ranks, the algorithm won't select such resource sets. In this paper we present a new resource selection algorithm based on clustering resources by domain to improve the old algorithm. We make a simulation experiment which uses Cactus application as specific application, experimental results show that new algorithm can select the resource set which match user's resource requests better.In 2004, Harbin Institute of Technology presents a Cluster-based Grid Resource Selection algorithm (CGRS algorithm) for synchronous iterative applications. Using a new resource cluster algorithm, CGRS algorithm can decrease the number of the resource sets that should be ranked. But the algorithm only simply uses stationary value to evaluate the link performance. In this paper we improve the CGRS algorithm, and then make a simulation experiment to analysis the stationary value quantitatively and compare the performance of the improved algorithm and the old one.At last the paper introduces Grid fault-tolerant technology, and then proposes a fault-tolerant resource selection method according to the characteristics of Grid and the technique of primary/backup copies.