An Enhanced Data-Locality-Aware Task Scheduling Algorithm for Hadoop Applications

Abstract

In general, Hadoop improves the task scheduling performance by determining data locality based on the location in which the input splits and MapTask are executed. However, if an input split consists of multiple data blocks that are distributed and stored in different nodes, this data location method fails to cope with the degradation in processing performance due to the increased frequency of data block copying. We propose a task scheduling algorithm that solves this issue by defining a method to classify data locality taking into account the location of all data blocks that comprise an input split, categorizing tasks based on the defined method, and sequentially assigning tasks according to a given priority. This study measures the performance of the proposed algorithm through a comparison of the total processing time, MapTask performance time, and data block copying frequency between the proposed algorithm and Hadoop's default task scheduling algorithm. The test results show that the proposed algorithm improved the total processing time by up to 25% and the data block copying frequency by up to 28%, when compared to the default algorithm.