The anatomy study of high performance task scheduling algorithm for Grid computing system

Large-scale computation is frequently limited to the performance of computer hardware or associated cost. However, as the development of information and network technologies thrives, idle computers all over the world can be utilized and organized to enhance overall computation performance; that is, Grid environments that facilitate distributed computation. Hence, the dispatching and scheduling of tasks should be considered as an important issue. Previous studies have demonstrated Grid environments that are composed of idled computers around the globe and are categorized as a type of Heterogeneous Computing (HC). However, scheduling heuristics currently applied to HC focus on the search of minimum makespan, instead of the reduction of cost. In addition, relevant studies usually presume that HC is based on high-speed bandwidth and the communication time is ignored. Further, in response to the call for user-pay policy, as a user dispatches a job to a Grid environment for computation, each execution task would be charged. It is difficult to estimate a job will be dispatched to which and how many computers; it is impossible to predetermine scheduling heuristic which is proposed in previous studies will result in the optimal makespan, and mention actual cost and risk. Therefore, this study proposes ATCS-MCT (Apparent Tardiness Cost Setups-Minimum Completion Time) scheduling algorithm that composes of execution time, weight, due date, and communication time factors to testify that the ATCS-MCT scheduling algorithm not only achieves better makespan than Min-min scheduling heuristics do but also reduces costs.