RB+-Tree and Its Implementation in Column-Stored Data Warehouse

The index technology is one of the key issues to improve the efficiency for massive data queries. Traditional index technologies such as B⁺-tree index has achieved good performance in the update transaction environment. However, the performance remains dissatisfactory by directly applying it to column- oriented analytical data warehouse. In this paper, we propose a novel tree-based index: RB⁺-tree (Reduced B⁺-tree) by analyzing the characteristics of column-oriented data warehouse query environment. The RB⁺-tree, with the improved structure and the bottom-up index creation approach, greatly improves the performance of create and search operations as well as the space usage. Further more, we apply RB⁺-tree to column-oriented data warehouse, and construct the rowid-index and value-index. Especially, we propose the join-index for multi-table join based on the star schema, which improves the tuple reconstruction and multi-table join performance effectively in column-oriented data warehouse system. The experimental results on the data warehouse benchmark data set SSB verify the effectiveness of the proposed method.