Speed-based Load Balancer for Scheduling Reduce Tasks to Process Intermediate Data of MapReduce Applications on Cloud Computing
Tzu-Chi Huang, Kuo-Chih Chu, Ce-Kuen Shieh, Ming‐Fong Tsai · 2015
MapReduce is a programming model used to develop applications on cloud computing. However, MapReduce strongly relies on the runtime system to handle issues of task managements in order to get a better performance. While most research works focus on scheduling Map tasks to improve performances, MapReduce is identified by this paper to have the potential for performance improvements through the Speed-based Load Balancer (SLB) for scheduling Reduce tasks. According to observations on experiments of Inverted Index, Radix Sort and Word Count, MapReduce can use SLB to outperform the native scheduler used by Hadoop in the runtime system.