Optimizing Parallelism of Big Data Analytics at Distributed Computing System

Abstract

Since advent of information revolution, there have been a lot of interest at big data analytics as well as big data. In the big data analytics, it is essential that not only extracting valuable information from the big data but also processing the data rapidly. Therefore, the distributed computing systems which process the analytics concurrently with parallel programming model based distributed processing framework as well as provide data analytics related libraries get attention of researchers. Several big data analytics programming models are studied that implemented for processing and generating huge data sets. However, developing the big data analytics in the distributed computing systems with utilizing parallel processing framework needs expertise in each area. In this paper, we demonstrate there is huge gap among usages of processing units if the big data analytics are naively executed at the distributed system. And we also prove that applying proper parallelism of those methods results in 1.5 to 3.3 times improvement of execution time compared to default parallelism.