Estimating job execution time and handling missing job requirements using rough set in grid scheduling

Abstract

Efficient scheduling of jobs in grid environment is a challenging task. To perform better resource utilization and proper resource allocation, the factor job runtime is essential. Accurate estimation of runtime helps to reserve resources in advance, provide user level QoS. But it is difficult to estimate the runtime of data intensive applications. Users are required to provide the runtime estimate of the job, but the user given estimates are inaccurate leading to poor scheduling. In this paper, we have used rough set techniques to analyse the history of jobs and estimate the runtime of the job. This requires maintaining a history of jobs that have executed along with their respective runtime. Our proposed rough set engine groups similar jobs and identifies the group to which the newly submitted job belongs. Based on this similar group identified, the runtime is estimated. Mostly users are not aware of resources, submitting incomplete job requirements. These missing job requirements affect data analysis. Those missing values should be accurately predicted. Missing value handler designed using rough sets fills the most probable value for missing attributes and then the runtime is estimated.