The Research Of A Trust Model-Based Task Scheduling Algorithm For Data Grid

Task scheduling is a key technology in grid computing; it can optimize the use of resource, improve the utilization of resource and provide users with better quality of service. In grid, especially in data grid, establishing trust model is considered to be an important measure of ameliorating the grid security.The current task scheduling model and algorithm are mainly aimed at computing grid, without considering data transmission overhead, while in data grid, transmission overhead sometimes may exceed computing overhead, which seriously affects scheduling performance. Therefore, people urgently need a task scheduling model and algorithm which are suitable for data grid. According to the problem above, this paper proposed a novel scheduling model that is suitable for data grid task scheduling, and based on trust mechanism, improved the Min-Min algorithm.The main work and innovation are as follows:(1) The scheduling model proposed in this paper separates traditional single scheduler into task scheduler and data scheduler. Task scheduler is in charge of task scheduling, data scheduler is responsible for data file selecting and scheduling, data scheduling assists task scheduling and executing. Data scheduler works in parallel with task scheduler, which decreases scheduling time.(2) This scheduling algorithm first establishes trust model for every file server and calculates its trust value; then in selecting task-required data file, it considers file server's trust value and data transmission time, and selects the node with bigger trust value and smaller data transmission time; in data transmission time calculating and transmission path selecting, it adopts the shortest path-Dijkstra algorithm, that is, selecting the path with the shortest transmission time between file server node and task execution node as file transmission path.(3) We used GridSim to verify the improved Min-Min task scheduling algorithm, the experiment results shows that although this scheduling algorithm expands the task completion time, the success rate of task execution is apparently raised.