Allocating Resource in Grid Workflow Based on State Prediction

To the workflow scheduling of DAG-based workflow in which time constraints are rigid, a new algorithm based on the model of resource state reliability was put forward. According to the requirement of the workflow execution time submitted by users, the algorithm found critical paths and obtained the probability for all resources being the idle state at the time when the task arrives by using Chapman-Kolmogorov backward equations, and then made the choice of the resource system scheme with the lowest expenditure under the certain credit level of the state reliability for the resource combination on the critical path in DAG. The simulation shows the validity of algorithm.