Research On Task Scheduling Algorithms Of Grid Computing Based On Multiple Qos Constrain

In recent years, Grid Computing has become one of the most popular technologies of Internet. Grid Computing is one of the high performance computing which combites the distributed computing and the traditional high performance computing. It can conquer the difficulty of management under the heterogeneous environment, and collaborate multiple QoS(Quality of Service) constrains. Grid Computing takes the advantage of many idle computers as its own resources and uses them to compute some large-scale task like prediction the weather, prediction the track of missile.etc efficiently.Amount the most hotest point researches of Grid Computing research area, task scheduling algorithms is very important. The quality of task scheduling algorithms determines the quality of the whole Grid Computing environment because there are so many different traits among workstations, namely performance, Operation System, environment.etc, so design a good algorithms to solve the above question is the primary task.Research results show that task scheduling algorithm is a NP-Hard question. It's hard to get the best solution to this kind of problem, even there isn't the best solution to them. So finding the satisfaction solution instead of the best one is the main task what this thesis focuses on.This thesis describles the development of the Grid Computing and researches the problem of multiple QoS constrain in the task scheduling environment deeply. A task scheduling model under multiple QoS constrain with user satisfaction in the Grid Computing Environment is designed.The first task is putting this model into Max-Min, and Min-Min algorithms. This thesis proposes a method using the absolute number of the margin between the predicted satisfaction value and the user's satisfaction value instead of the minium makespan or maxium makespan. Then analyze for the result of the experiment and compare the defference between the model has proposed in this thesis and traditional Max-Min and Min-Min algorithms is performed. The second task is the implement MQoS-GA which contains this model and based on traditional Genetic Algorithms. The final task is analyzing the result between different initial No. of population.There are so many QoS constrain specifications in the Grid Computing environment, but the author takes the makespan, priority and credit as the most important specifications and uses them as the multiple QoS constrain. Giving the satisfaction function is assigned to each index and a synthesical satisfication function is constructed according to these index. And then the task scheduling algorithm is designed and implemented based on satisfication optimization model.The simulation results show that this algorithm can collaborate multiple QoS constrain efficiently and take the advantage of Genetic Algorithms to solve multiple target optimization problem under the requirement of user's satisfaction.