Hadoop Task Scheduling - Improving Algorithms Using Tabular Approach

Abstract

Map Reduce is a widely adopted implementation in many fields like that of scientific analysis for data processing, processing data on web as well as areas like high performance computing.Computing systems with heavy data handling requirements should provide an effective scheduling method so that utilization is enhanced.The major problems encountered in scheduling MapReduce jobs are mostly caused by locality and overhead of synchronization.Various other factors like fairness constraints and distribution of workload have been discussed further in the paper and are the highlight of the paper.The paper describes the Hadoop and working of MapReduce in brief.Our paper compares different scheduling methods for handling the mentioned issues in MapReduce and they are compared on the basis of their strength, weakness and features.Through this paper, we aim to consider three different factors along with introducing a small modification to enhance the scheduling by using tabular approach.The purpose is to provide researchers further with a direction in which they can proceed and come up with a more generic algorithm for task scheduling in Hadoop MapReduce.