Performance Evaluations of Two Join Methods: Join Core and Join Indices

Author

Reham Almutairi, Southern Illinois University CarbondaleFollow

Date of Award

12-1-2017

Degree Name

Master of Science

Department

Computer Science

First Advisor

Hou, Dr. Wen-Chi

Abstract

Traditional join algorithms can be categorized into three groups: hash-based join, sort-merge join, and nested-loop join. Many variations of these algorithms have been proposed during the last few decades to improve the performance. In this paper, we have compared between two of the most efficient join methods: Join indices and Join Core. Join indices generate index tables that contain tuples identifiers for matched tuples. It scans each input relation only once, the join index once, and scans temporary files twice. On the other hand, Join Core is a data structure created to facilitate complex join queries promptly. With the Join Core, join queries can be handled quickly without performing costly join operations. Also, no intermediate results need to be retrieved during the runtime. Consequently, join queries can be answered rapidly. We implement the multi-way Jive-join version of the Join indices, using Java and TPC-H benchmark datasets. Our experimental result shows that even multi-way Jive-join method has better memory utilization, it takes a long time to process queries since it requires generating output files and temporary files before generating the final results. However, our results show that processing queries with Multi-way Jive-join is faster than using MySQL.