Techniques for semi-automatic generation of data cubes from star-schemas

Abstract: The aim of this thesis is to investigate techniques to better automate the process of generating data cubes from star- or snowflake schemas. The company Trimma builds cubes manually today, but we will investigate doing this more efficiently. We will select two basic approaches and implement them in Prototype A and Prototype B. Prototype A is a direct method that communicates directly with a database server. Prototype B is an indirect method that creates configuration files that can, later on, get loaded onto a database server. We evaluate the two prototypes over a star schema and a snowflake schema case provided by Trimma. The evaluation criteria include completeness, usability, documentation and support, maintainability, license costs, and development speed. Our evaluation indicates that Prototype A is generally outperforming Prototype B and that prototype A is arguably performing better than the manual method current employed by Trimma.