FIESTA: A Framework for Schema Evolution in Multidimensional Databases
Markus Blaschka · mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) · 2000
New application areas for databases like data warehousing and OLAP (Online Analytical Proc-essing) deploy the multidimensional data model in order to describe the application domain. Consequently, OLAP systems are represented by a multidimensional database schema to adequately reflect the application semantics. FIESTA presents a methodology for the evolution of such multidimensional schemas. Core of the thesis is a schema evolution algebra which comprehends a formal multidimensional data model together with corresponding schema evolution operations. Since OLAP systems are typically implemented as additional layer for relational database systems, the processing of sequences of schema evolution operations in a relational database system is presented. To this end, we formally describe how a multidimensional schema can be mapped to a corresponding relational database schema. In order to fully maintain the multidimensional semantics during this transformation, a corresponding meta schema is introduced as extension of the relational system catalogue. For a consistent processing of evolution operation sequences, a transformation to corresponding relational evolution commands is performed. These relational evolution commands adapt the relational database schema together with the instances and update the contents of the meta schema accordingly. A prototype for FIESTA has been implemented as part of a graphical design environment for data warehouses. In this environment, multidimensional schemas are presented at the user interface by means of a specialized graphical notation. This notation is an extension of the well-known Entity/Relationship approach. For internal processing the graphical representation is transformed to an algebraic description of the multidimensional schema.