Optimization And Research Of Scheduling Strategy On MapReduce Based Clouds

With the progress and development of computer technology,cloud computing as new computing model which developed from distributed computing, grid computing and parallel computing has been rapidly development.C loud computing controls all computing resources and storage resources from the cloud, so as to provide cheaper and more high quality service to users.Cloud computing as a commercial computing model, user ’s satisfaction to the service plays a decisive role to the development and operation of cloud computing. So the subject that worthy of further study is how to meet customers in the quality of service(QoS) at the same time maximize the system resource utilization.For the scheduling performance problem of the existing MapReduce architecture and user demand for quality of service problem, this thesis presents a cloud computing task scheduling strategy which based on the genetic and ant colony optimization algorithm to improve the QoS. The algorithm on the premise of guarantee user ’s QoS, at first take advantage of the genetic algorithm’s strong global search ability to find the optimal solution of the scheduling problem, then converts the optimal solution to the initial pheromone of ant colony optimization algorithm,finally after information communication and positive feedback to find the global optimal solution of the task scheduling problem.Based on the analysis of the existing scheduling algorithm,for the traditional scheduling algorithm fail to reduce the average completion time and the lack of load balancing, this thesis proposes a scheduling strategy which based on the shortest completion time and attemps to assigning the task to the suitable resource in order to reduce the aver age job completion time.Meanwhile,for the Master node as the scheduling center node under heavy load in the existing cloud system,and based on the shortest completion time scheduling strategy,this paper presents a scheduling model which has new Scheduler Master node. The SchedulerMaster node is responsible for scheduling tasks and resources by matching calculation,so it can effectively reduce the calculation load of the Master node.Finally,in the C loudSim cloud computing simulation platform,by evaluating the average job execution time and user QoS,the effectiveness and stability of the presented scheduling algorithm and scheduling model is verified.