Preemptive ReduceTask Scheduling for Fair and Fast Job Completion
Yan‐Dong Wang, Jian Rong Tan, Weikuan Yu, Li Zhang, Xiaoqiao Meng, Xiaobing Li · International Conference on Autonomic Computing · 2013
Hadoop MapReduce adopts a two-phase (map and reduce) scheme to schedule tasks among data-intensive applications. However, under this scheme, Hadoop schedulers do not work effectively for both phases. We reveal that there exists a serious fairness issue among jobs of different sizes, leading to prolonged execution for small jobs, which are starving for reduce slots held by large jobs. To solve this fairness issue and ensure fast completion for all jobs, we propose the Preemptive ReduceTask mechanism and the Fair Completion scheduler. Preemptive ReduceTask is a mechanism that corrects the monopolizing behavior of long reduce tasks from large jobs. The Fair Completion Scheduler dynamically balances the execution of different jobs for fair and fast completion. Experimental results with a diverse collection of benchmarks and tests demonstrate that these techniques together speed up the average job execution by as much as 39.7%, and improve fairness by up to 66.7%.