An efficient query processing optimization based on ELM in the cloud

Abstract

Nowadays, MapReduce has emerged as a facto programming model for parallel processing of large-scale datasets with a commodity cluster of machines. MapReduce and its variants have been widely researched in the industry and academic communities. ComMapReduce further extends MapReduce by adding lightweight communication mechanisms and also enhances the efficiency of query processing applications. However, we find that the performance of query processing applications changes a lot in different communication strategies of ComMapReduce framework. It is necessary to identify the most optimal communication strategies of the query processing applications. Extreme learning machine (ELM) can exactly provide classification performance with an extremely fast training speed. Therefore, in this paper, first, we propose an efficient query processing optimization approach based on ELM in ComMapReduce framework, named ELM_CMR. Then, we design two implementations of our ELM_CMR approach to further optimize the performance of query processing applications. Finally, extensive experiments are conducted to verify the effectiveness and efficiency of our proposed ELM_CMR.