Automatic Selection of Bitmap Join Indexes in Data Warehouses Using CFPGrowth++ Algorithm

Abstract

In the context of complex data warehousing, Typically, the analysis and decision-making process for Data Warehouses schematized in a relational star model is conducted through OLAP (On-Line Analytical Processing) queries. These queries are generally complex, characterized by several operations of selections, joins, grouping and aggregations on voluminous tables. Which requires a lot of computing time and therefore a very high response time. The cost of running OLAP decision queries on large tables is very high. The reduction of this cost becomes essential to allow decision-makers to interact within a reasonable time frame. The objective of this study is to enhance system performance by minimizing the response time of OLAP decision-making queries. The approach proposed in this article aims to search for frequent patterns for the automatic selection of binary join indexes used for reducing the execution costs of OLAP decision-making queries. To automatically generate the configuration of binary join indexes minimizing response time, an implementation of the CFPGrowth++ frequent pattern matching algorithm was well carried out and then applied to a load of queries on a test Data Warehouse created using the Analytical Processing Benchmark 1 (ABP-1) test bench, in order to validate our approach. The results of the experiment indicate that the index configuration produced by the proposed approach leads to a significant improvement in performance improvement of approximately 75%. We note that for a large portion of the load, execution time is significantly improved after applying our approach. The overall query execution time decreased compared to the general context. The overall execution time for queries decreased from 20,032.57 seconds before the application of our approach to 5,388.49 seconds after applying our approach. The experiments carried out show that the index configuration generated by the proposed approach allows a very performance gain.