Tine Borovnik (2010) Modern approaches to data warehouse design. MSc thesis.
In this Master’s thesis data warehouses and approaches to their design are presented. On one hand, classical approaches, which have been in use for almost two decades, are still popular; on the other hand, new approaches are on the horizon, actually being used already, yet still undergoing an intensive development phase. A typical data warehouse architecture and its placement in the modern information environment is shown, followed by a description of a variety of data warehouses and the comparison between them. The framework for the next data warehouse generation DW 2.0 is presented. Dimensional models rank among the most widely applied data models. A detailed description of the way of such modeling in relational databases is shown. The use of materialized views is shown as a technique for performance improvement. Multidimensional databases are improving constantly, therefore the cubes as their basic component are more and more widely used. In addition to speed, their advantages in the field of advanced handling of hierarchies are presented. DW 2.0 framework foresees the building of data model on three levels, starting with an overall less detailed model and concluding with a detailed physical model. Anchor modeling represents a prospective modern approach to the data model design which is paving its way into data warehouses. It is based on the sixth normal form. Simple and clear structures for ensuring data quality, loaded by ETL processes are presented. A special chapter deals with a detailed comparison of approaches regarding the properties such as suitability to the types of data warehouses, ease of maintenance and usage, acceptance, history conservation etc. Presenting a solution example, another chapter includes a practical presentation of the use of the described approaches on an example of a data warehouse for a telecommunication company. Considering their advantages and disadvantages as well as the needs of such a company, among the described the most suitable approaches have been selected to be used for the development of the data model.
Actions (login required)