Research On Grid Resource Scheduling Techniques Based On Computational Economy Model

Grid resource management and task scheduling are very complex. Traditional resource management and task scheduling methods can not adapt to the grid environment. In recent years, to introduce the economics model into the grid system and apply the method of economics to the grid resource management and task scheduling is a hot spot in the field of grid computing.At present, many grid task scheduling algorithms which based on computational economy model have been proposed, however, most scheduling algorithms only focus on the single optimization objective, for example, the task's makespan or execution cost. Generally, users always hope tasks are able to execute in the shortest time and with the most economical manner. In order to meet these multiple QoS requirements which have conflict with each other, we need to search efficient and cost-effective scheduling algorithms that support multiple QoS requirements.In view of the problem of most of scheduling algorithms can not meet users'multiple QoS requirements, Two scheduling algorithms, called Classified Optimization Scheduling Based on Cost and Time Trade-off (COSBCTT) and Classified Optimization Scheduling Based on Multiple QoS (COSBMQ), which based on computational economy model are proposed in this dissertation. COSBCTT base on the idea of Min-Min algorithm, to classify resource for scheduling by improving MinCTT (Min-Min Cost Time Trade-off) algorithm. It is aim at minimum of tasks'overall makespan and execution cost. COSBMQ which base on COSBCTT synthetically considers three kinds QoS requirements:the makespan, execution cost and urgency of user task. COSBMQ has merits of COSBCTT, and giving consideration to the task urgency requirement while optimizing the makespan and execution cost. the task that have minimum trade-off value and high urgency will be mapped to corresponding resource at first. It is aim at decrease of tasks'overall makespan and execution cost, and increase the number of tasks executed within deadline.We design simulation experiments based on the simulation tool GridSim and compare our scheduling algorithms against MinCTT scheduling algorithm. Experimental results show that COSBCTT can achieve the least overall makespan and execution cost of tasks. COSBMQ that has better scheduling performance than MinCTT algorithm can achieve less overall makespan and execution cost of tasks and the highest percentage of tasks executed within their deadlines. In addition, the experiments prove that the trade-off factor has a great impact on the performance of MinCTT algorithm, however, corresponding trade-off factors have a small impact on the performance of COSBCTT and COSBMQ, it can achieve good scheduling when proper trade-off factors are choosed.