Research And Improvement Of Capacity Scheduling Algorithm In Cloud Computing

In recent years, as a new high-performance computing model, cloud computing has been focused by majority of researchers and scholars,even that many companies also have provieded their own platform, such as the Eucalyptus of University of California, Hadoop platform of Apache, as well as MongoDB of10gen ect. As an open source, Hadoop platform has been widely used. The advantages of this platform included in distributing, high efficiency, low costing and strong reliability etc. Job scheduling is one of the most important technologies of cloud computing, it is a key influence to the whole platform performance and resource utilization. The function of job scheduling technology is let jobs have a reasonable distribution and processing. Its goal is both to make the whole system can be run orderly, and use resources fully and effectively, and the same time, let customer’s satisfaction as high as possible. However, with the increasing demand of users, types and scales, to study a new job scheduling algorithm that not only meets the above requirements, but also combined with the practical application has a great significance.Currently, widely used job scheduling algorithm is FIFO, which is simple and has low cost, but it’s only suitable to meets of single job or small job. Fair Scheduling algorithm supports multiple users used to resources fairly, so its can meet a lot of jobs access into the system, but this can easily result in waste of resources; The capacity scheduling algorithm draws on the inadequate of fair scheduling, and allocate the resources bases on the job performance, but this allocation strategy is too simple to easy to fall into the local optimum. Some scholars start from the system resources, system configuration, operations etc. They are tried to propose some improved algorithm through deeply study.This paper in view of the system configuraiotn, from the total running time of the job, the average running time and waiting time, using of the simulated annealing algorithm in combinatorial optimization problem’s advantage which can avoid local optimal. So this paper combined with capacity scheduling algorithm, Use the Simulated Annealing to the optimization of the job scheduling, based on the capacity scheduling algorithm proposed a new algorithm, constructs a simulated annealing mathematical mode, and selects the default search strategy of capacity scheduling as the initial solution, proposes a new objective function, build the solution space of the job, and choose a logarithmic function as annealing strategy. This function aimed at improving the job operational efficiency and at the same time reducing the job’s waiting time. Simulated Annealing job scheduling algorithm has been improved in order to improve the learning speed, and add memory function; the algorithm can greatly reduce the number of iterations to improve the research speed and the convergence speed.This paper implements the algorithm under the Hadoop platform at last, which includes the configuration of the internal and the configuration of four scheduling algorithms. The improved algorithm and previous three algorithms are run into the platform, and get the total running times and waiting times. Finally, this paper proved the effectiveness of the improved algorithm via the comparison and analysis to the experiment results.