Scaling beyond one rack and sizing of hadoop platform

Abstract

This paper focuses on two aspects of configuration choices of the Hadoop platform. Firstly we are looking to establish performance implications of expanding an existing Hadoop cluster beyond a single rack. In the second part of the testing we are focusing on performance differences when deploying clusters of different sizes. The study also examines constraints of the disk latency found on the test cluster during our experiments and discusses their impact on the overall performance. All testing approaches described in this work offer an insight into understanding of Hadoop environment for the companies looking to either expand their existing Big Data analytics platform or implement it for the first time.