Research Of Job Scheduling Algorithms In Cloud Computing

Due to its characteristics, including the enormous processing capacity and storage capacity, high efficiency, virtuality and affordability, cloud computing, as an emerging compute model, has been greatly concerned by academics and merchant. Cloud computing System has the ability of processing large-scale dataset efficiently and concurrently, even it is deployed on inexpensive server. Thousands of nodes make up the cloud platform, therefore, how to adopt the appropriate task scheduling technology to process various tasks cooperatively is becoming particularly important. A better job scheduler can achieve lower task response time and system throughput, as well as higher utilization ratio of system resources. Hence, the research of task scheduling strategy has important implications for the development of cloud computing.On the basis of an in-depth study of the concept, service pattern, characteristics and development prospect of cloud computing, systematic learning of Hadoop platform, detailed discussion of HDFS and MapReduce, this paper proposes the existed inadequate points and modification plan of three common job scheduling algorithms, that is FIFO, Fair Scheduler and Capacity Scheduler, laying the theoretical foundation for following research.Aiming at shortening total and average completion time of tasks, a novel simulated annealing algorithm combining capacity scheduler is proposed. The novel algorithm adopts SA scheduler for global search and adding memory parameters to choose tasks instead of the original simple strategy, which would effectively decrease iterations, increase search speed and convergence rate. Eventually, output the global optimal solution.We could conduct the modified simulated annealing algorithm experiment with simulation platform CloudSim, and compare the performance with the job scheduling algorithms job scheduling algorithms. The experimental result shows that the modified algorithm can achieve higher effectiveness and better optimization.