Task Selection for Scheduling using Hadoop Scheduler

Abstract

MapReduce is a prevalent model for data intensive applications. This covers the difficulties of parallel programming and provides an abstract environment. Hadoop is a benchmark for Big Data storage by being able to provide load balancing, scalable and fault tolerance operation. Hadoop output is mainly dependent on scheduler. Various algorithms for scheduling [6-10]have been suggested for various types of environments, applications and workload. In this work new task selection method is developed to facilitate the scheduler, if a node has several local tasks. Experimental result shows an improvement of 20% in respect of locality and fairness.