The Research On Task Scheduling Strategy In Cloud Environment

Cloud computing is a new computing model after grid computing, distributed computing and parallel computing. It has changed the traditional service model, by combining with the Internet, providing a new business model. Current cloud computing environment resources are mainly virtual machine resources, in particular through the use of virtualization technology to virtualize various hardware of the physical resources, forming a virtual resource pool dynamically deploy virtual machines to user for using transparently. With the data size of the cloud continues to expand and the number of users continues to increase.How to allocate and utilize the resources of cloud environment rationally and schedule the massive task of user-submitted efficiently, so that users do not wait too long and be able to better meet the user quality of service(Qo S), is an important issue in the study of cloud computing technology. Thus, the task scheduling strategy research has important implications for the development of cloud computing, but also the challenges currently facing.For these reasons, the paper detailed analysis of cloud computing research background, meaning, current research as well as task scheduling algorithm on Hadoop platform. About task scheduling strategy for cloud, I have researched as followings:Firstly, this paper analyzes the background of cloud computing and cloud computing task scheduling research status and systematicly elaborates concept of cloud computing, the key technology, Map Reduce V2(YARN) distributed programming frameworks such as cloud computing technologies. Summarize the existing cloud task scheduling algorithm.Secondly, the genetic algorithm(GA) with parallel and global search, solving task scheduling problem in cloud can get better results. By analyzing the advantages and disadvantages of traditional genetic algorithm, this paper puts forward Dual Fitness Multiplied Genetic Algorithm(DFMGA), this algorithm considers the average time task execution, thus ensures the user’s overall satisfaction.Finally, Haoop as open source cloud platform, the industry and academia pay much attention to this for research and applications. In this paper, Hadoop is as the analysis and verification platform, the paper studies the task scheduling implementation of it, analyzes the three commonly used scheduling method s in Hadoop: FIFO, fair scheduling, capacity scheduling. This paper, through comparing the performance of DFMGA and traditional algorithms, gives the performance of FIFO, traditional genetic algorithm and improved genetic algorithm DFMGA in total task completion time, the average time to complete of these two aspects, demonstrates improved genetic algorithm(DFMGA) having shorter execution time and less cost consuming.