Stochastic model and evolutionary optimization algorithm for grid scheduling

Grid computing deals with computationally intensive distributed resources on heterogeneous environment, so grid scheduling is a fundamental challenge and is critical to performance and cost. Traditional grid scheduling algorithms most use deterministic models. But grid environments in the real world are subject to many sources of uncertainty or randomness, such as network status, job execution costs, which are often not known precisely in advance. A good model for a scheduling problem should address these of uncertainty. This paper presents a new stochastic model for grid scheduling and a novel evolutionary scheduling algorithm based on this model. Furthermore the optimization methods are used to improve grid QoS. At last we demonstrate the grid workflow management architecture on which the solution can be practically performed. The simulated experiments show that our scheduling algorithm is feasible.