|
The development of internet technology has caused grid being born as a distributed computing platform, it mainly provides resource sharing and collaboration, these resources include computing resources, storage resources, information resources, hardware resources, software resources, etc. It is the combination of all the resources on the network together to form a supercomputer to carry out work, thereby eliminating the resource island.In the dynamic, heterogeneous grid environments, task scheduling is one of the main contents in grid computing research, and it is a complex and challenging issue. Based on the analysis of the basic task scheduling algorithm, we focuses on the classical Min-Min scheduling algorithm, and points out the advantages and disadvantages of the scheduling algorithm, the advantage is the ability to ensure the minimum total execution time, and the drawback is that the load is not balancing, and can not be achieved from an economic point of view. For both load balancing and economic aspects of the Min-Min scheduling algorithm is proposed: the price mechanism based on the improved Min-Min scheduling algorithm and cost-based the improved Min-Min scheduling algorithm.Experiments show that the improved algorithm balances the load and saves part of the cost in a certain extent.Firstly, the development of the grid background was analyzed, including domestic and foreign situation of the latest research, and the concept of grid, features and applications were introduced, showing the significance of this study. On the basis of the grid, the related concept of grid task scheduling was drawn, and the characteristics, objectives, classification and the three models of grid task scheduling were analyzed, and four metrics were proposed in grid task scheduling algorithm: the total execution time, quality of service, load balancing, total economic costs.Secondly, the Min-Min scheduling algorithm’s advantages and disadvantages were focused on. the seven classic common grid scheduling algorithms were analyzed and the advantages and disadvantages of each method were obtained, through analyzing the Min-Min scheduling in detail, the advantages and disadvantages of the algorithm were illustrated.The advantage is the ability to ensure the minimum total execution time, and the disadvantage is load imbalance, the high cost and that the algorithm can not be achieved from the perspective of quality of service scheduling.Thirdly, two improved algorithms based on the price mechanism and cost-effective were proposed to improve the Min-Min scheduling algorithm. Elaborated the idea of the new algorithm, and through a mathematical model of the way, a detailed description of the algorithm, theoretically we proved that the improved algorithm were greatly improved in terms of load balancing and cost.Finally, two new algorithms were verified on grid simulator. The reasons for using grid simulator were first explained, five common grid simulators were described and the differences between them were said, the reasons for selecting Gridsim simulator were explained, and a detailed description of the simulator, including simulator configuration, installation and operating environment, inside the simulator and using classes were introduced. According to simulator experiments, we put forward the data and comparison chart, and analyzed the reasons for the experiment that the new algorithm could improve the efficiency of the original algorithm, and improved the overall performance of the network, obtaining satisfactory results.
|