A Smart Agent-Based Grid Computing Platform

Most of Grid computing platforms are usually configured with high performance computing servers, supercomputers, and cluster systems. The research is about to design an aggressive Grid computing platform by utilizing a large number of pervasive PCs as a Grid component. For this goal, an effective configuration method to construct a group of PCs is required as a single Grid component to be used in the same way as those cluster systems are designed and utilized as the same system technology. The configuration method is designed as a layered service architecture, i.e., physical information service, mobile management service, and point-based scheduling service, based on agent technology. The physical information service is to gather and maintain the status information of physical nodes, the mobile management service to guarantee the effective resource management by using the mobile agent, and the point-based scheduling service to provide a simple scheduling policy for the PC-based computing resource. Through this configuration method, the efficiency of the resource management and system throughput are expected to be increased. The experimental result shows that the system using this configuration method can support more than 90% of the expected performance given by a chosen set of running PC resources in general computing environment.