Resource Availability Evaluation in Service Grid Environment

Abstract

In dynamic grid environment, prediction and evaluation of resource availability are the prerequisite for reasonable resource selection and good QoS guarantee. Based on some related resource and task historical traces, probability theory is applied to resource availability prediction and evaluation. The availability metrics including resource off-line time, local task execution time, waiting queue length and waiting time are presented and the distribution functions of these metrics are given and proven. The experiment results show that the prediction is effective, and the amount of candidate resources determined by resource availability evaluation is decreased significantly, therefore lowering time complexity of task scheduling.