Dimitar Mišev

• Array databases are a quickly expanding category of database management systems that treat large multidimensional arrays as first-class database citizens. Array data itself is almost always linked to additional, non-array information, but this is not adequately handled in today’s systems. Array databases specialize in the management of array data, while other systems, e.g. relational DBMS, have at best only very basic support for arrays. As a result, handling array data in practice most often requires either multiple DBMS with manual data integration and synchronization, or dedicated solutions constrained to a narrow domain. This thesis addresses this situation by extending the relatonal data model with support for multidimensional arrays in a non-intrusive way that is orthogonal to its set semantics. The array model itself is declarative, optimizable, and minimal, yet powerful enough for application domains in science, engineering, and business. The algebraic formalization is materialized into an official standard ISO SQL extension known as SQL/MDA. A proof of concept mediator implementation of SQL/MDA – utilizing a new array processing engine optimized for modern hardware and a standard relational DBMS – demonstrates practical feasibility of the established concepts. All in all, this thesis covers in completeness the topic of array / relation integration in databases and presents a theoretically sound and practically viable solution.