Binary Information Press
MapReduce is a well-known programming model proposed by Google, and is implemented as open source software called Hadoop, which is now-a-days widely used to process large-scale data. Some schedulers for Hadoop mainly care about the resource utilization but not the constraint of the completion time. Some schedulers focus on this constraint. However, these schedulers do not solve the issues of remaining time prediction and task exceeding. In this paper, the authors propose a scheduling strategy to ensure that each job could complete in time so as to meet the performance goal.