A New Scheduling Algorithm to Reduce Computation Time in Hadoop Environment

Nowadays, the Hadoop open-source project with the MapReduce framework has become very popular as it processes vast amounts of data in parallel on large clusters of commodity hardware in a reliable and     fault-tolerant manner. MapReduce was introduced to solve large-data computational problems, and is    dependent on the divide and conquer principle. Time and scheduling are always the most important aspects, hence in the past decades in the MapReduce environment, many scheduling algorithms have been proposed. The main ideas of these algorithms are increasing data locality rate, and decreasing response time and completion time. In this research we have proposed a new hybrid scheduling algorithm (HSMRPL) which uses dynamic job priority and identity localization techniques, and focuses on increasing data locality rate and decreasing completion time. We have evaluated and compared our algorithm with hadoop default schedulers by running concurrent workloads consisting of the WordCount and Terasort benchmarks. The results show that our proposed algorithm has increased the localization rate by 10.4% and 18.5% and the speed by 3.14% and 3.3% compared to the FIFO algorithm and the Fair algorithm respectively.