Information Service Prototype System for Run-time Prediction of Grid Applications

In order to improve the utilization of grid resources and execution performance of applications, online runtime prediction of applications is needed as direction for job scheduling system or grid users. In this paper, we build up an information service prototype system in which prediction information is periodically generated, then formatted and registered into grid information service. We use resource mapping methods in predicting run-time and also run a series of process to test the prediction performance of our prototype system. The experimental results show that the prototype system is sensitive, in low overhead and can give run-time prediction in acceptable accuracy.