VPRS-based self-study and prediction techniques of grid system

The architecture of self-study and prediction system of computing grid is presented. The variable precision rough set and inconsistent decision information system theory is adopted to reveal the knowledge about the applications attributes, dynamic grid resources, and the applications performance. To identify the important attributes related to resource requirement and performance of grid applications, an incremental knowledge reduction technique is proposed. Using the discovered knowledge, the resource requirement and performance of grid applications can be predicted. The experiments show that the mean prediction error of the applications resource requirement and performance is less than 30%.