Study On Modeling Large-Scale Grid Platforms And Scheduling Algorithms

With the development of software and hardware technology, it is possible to dispose parallel applications on the sets of extensive distributed resources that increase constantly. And with grid's rising, that researching on the scheduling algorithms of these applications is an active area. In this paper, we have a research on modeling large-scale platforms and propose Mended Min-Min scheduling algorithms based on this model.There are there aspects of networks that must be taken into account in platform models: network latencies, bandwidth sharing and network topology. In this paper, we discuss these three aspects and propose a simple model of large-scale grid platform. We also study several classic scheduling algorithms and propose Mended Min-Min scheduling algorithm based on Min-min algorithm. This algorithm initiate the schedule with Min-Min scheduling algorithm, and then schedules the other meta-tasks according to the minimum transfer time of Data transfer subtasks. Lastly, we simulate the model of grid platform and the Mended Min-Min scheduling algorithm, and make a comparison with Min-Min scheduling algorithm. To the schedule of a large number of tasks, the result of this experiment shows that our algorithm achieves a higher performance than Min-Min algorithm because of taking the three characteristics of grid into account.