DefinePK

Abstract

MapReduce is a popular large-scale data-parallel processing model for analysing and processing large massive data sets. Its success has stimulated several studies of implementing MapReduce on Graphic Processing Unit (GPU). Hadoop’s has motivated research interest and has led to different modifications as well as extensions to framework. The Graphics Processing Units (GPU) are widely used in the High-Performance Computing world to enhance job throughput, as its architecture is quite data-parallel friendly. The problem is to find Software and/or hardware solutions that allow firms to discover, evaluate, optimize, and deploy predictive models by analysing big data sources to improve business performance or mitigate risk in major data processing tools. MapReduce has also been widely adopted to solve Big Data problems and in this scenario combination of CPU and GPU will provide huge advantage over the scheme where only CPU is utilized. By default, Hadoop supports some fairly simple scheduling policies e.g. FIFO, fair scheduling, or capacity scheduling. In this paper we will study about MapReduce implementation of big data analysis using heterogeneous CPU-GPU scheduling to improve the performance of the system.