Selection of Bitmap Join Index: Approach Based on Minimal Transversals

摘要

Decision systems handle a large volume of data usually stored in a data warehouse. The latter is modeled by a star schema that typically has a central fact table and a set of dimension tables. The queries corresponding to this type of model are very complex. In order to reduce the cost of running these queries, one common solution would be to ensure a good physical design of data warehouses. In this respect, the binary join indexes (BJI) are very suitable for reducing the cost of running these joins. In this paper, we introduce a BJI selection approach based on a key notion of the theory of hypergraphs, namely minimal transversal. The final configuration obtained is composed of several indexes that optimize the cost of executing of the set of queries. The carried out experiments prove the relevance of our approach versus those of the literature.