Energy-Aware Task Scheduling of MapReduce Cluster

Abstract

Energy consumption in data center is gradually exceeding other operating expenditures. It is imperative to enhance the energy efficiency of data center. In this paper, two heuristics on AIS (All-In Strategy) - TSA (Task Scheduling on AIS) and TSAGT (Task Scheduling on AIS of Global Tasks) are constructed based on job performance, data locality and resource utilization for energy-aware task scheduling. Priority queue is obtained according to the number of allocated slots of jobs within deadline. Resource utilization and data locality are considered for task scheduling. Task adjusting for minimizing the completion time of cluster was proposed, which assigns the task to the server with the remaining running time similar to its processing time. Experimental results show that, the performance of TSA and TSAGT are better than existing algorithms on the completion time of cluster. Specially, TSA outperforms TSAGT in effectiveness with less completion time and TSAGT has less cost than TSA.