Evaluating the Impact of Data Modeling on OLAP Applications using Relacional and Columnar DBMS

Abstract

Data Warehouses has consolidate as the decision support technology used by Organizations that uses OLAP applications to access the stored data. As these data volume increases more efficient approaches to process them are needed. To do so, both traditional relational databases management systems and columnar ones can be used, each one with their advantages over the Data Warehouse modeling. More normalized models are traditional among tuple oriented relational databases, whereas denormalized ones bring a better performance in columnar DBMS. A comparative study between MonetDB and PostgreSQL DBMS using TPC-H as a benchmark is presented here, to investigate which one is indicated to manage a Data Warehouse in information access. The results confirmed that, isolated, in denormalized environments MonetDB excels, while PostgreSQL is better for normalized modeling. In general, MonetDB stands out compared to PostgreSQL, with performance gains of almost 500% on normalized model, and over 1000% on the denormalized one.