LSP.1. Flexible Multidimensional Data Processing

Traditionally data management solutions are designed to work on well-modeled data schemas, usually developed in a first phase of the overall system design. For example, application requirements coming out of the business context are reflect in UML or ER-diagrams and then transferred into a database model (e.g. the relational database model). In a final phase, the database model is restructured according to well-known normalisation schemas and implemented within the database system. If changes are necessary, the whole deployment cycle potentially starts from the very beginning. However, in the context of a flexible BI environment, this traditional setup does not work any longer: deployment cycles are too complex, the final schema may not be known in advance, or the change of a schema is the norm, not the exception. Data management solutions have to be prepared and tackle this challenge. Within this topic, on a system architecture level, different processing and storage structure schemes have to be classified and discussed with special emphasis on the targeted application scenario. In a second step, flexible record structures and compression schemes have to be developed and experimentally investigated. Issues like performance behavior or storage consumption have to be measured and compared against each other. On a modeling level, the topic is supposed to develop extensions to the process of normalisation to be able to constraint the flexibility of a table. In summary, the topic is positioned to lay the foundations for an efficient flexible data management layer, following the principle "data comes first, schema comes second" which is especially true to flexible ad-hoc BI scenarios.

Main Advisor at Technische Universit├Ąt Dresden (TUD)
Co-advisor at Universitat Polit├Ęcnica de Catalunya (UPC)