Much research has been devoted to designing efficient join algorithms in the last few decades. In this paper, we compare two fastest join methods: Join indices and Join Core. Join indices generate index tables that contain tuples identifiers for matching tuples. Joins can be performed by scaning each input relation only once. On the other hand. Join Core is a data structure that stores join relationships to facilitate join query processing. With Join Core, join queries can be answered without having to perform costly join operations. We have implemented both methods and performed extensive experiments on TPC-H benchmark datasets and queries. Our experimental result shows that while both methods are much faster than conventional systems, such as MySQL. Join Core is the fastest.
展开▼
