Optimizing Performance of Hadoop with Parameter Tuning

¹ Faculty of Information Technology, Beijing University of Technology, Beijing China
² NO.10 Building, Suzhou Software Park, No.78 China Mobile Software Technology Co., Ltd, Suzhou China

^a chenxiang1@emails.bjut.edu.cn
^b yliang@bjut.edu.cn
^c liguangrui@cmss.chinamobile.com
^d ccheng@emails.bjut.edu.cn
^e liusifen@emails.bjut.edu.cn

Abstract

Optimizing Hadoop with the parameter tuning is an effective way to greatly improve the performance, but it usually costs too much time to identify the optimal parameters configuration because there are many parameters. Users are always blindly adjust too many parameters and are sometimes confused about which one could be changed at a higher-priority. To make optimization easier, classifying the parameter based on different applications will be helpful. In this paper, we will introduce a method that can classify these parameters in order that users can optimize performance more quickly and effectively for different applications.