RADD, raddstar: a rule based database schema ; compiler, evaluator, and optimizer

Database design typically results in SQL create table commands and integrity constraints which are specified by foreign-keys, cascading deletes, etc. Specific products of the important database vendors provide a limited form of higher level design of the database relations and their associations. The design is then very close to physical design of the schema which is implemented under the associated DBMS. This design and specification approach has a rather limited scope and is difficult to handle. And, data profiles– such as the velocity, priority, and frequency of transactions, or the tuple numbers of relations –are also not considered by this traditional approach to database design and database application development. This way, operational behavior is not considered and can not be improved during database design. Recent information system development approaches, starting with the design of the logical database schema by means of the database design tool of the DBMS vendor and then constructing the database applications with the help of a Web application builder, also do not consider these operational behavior details. Rather, they restrict the designer to construct the schema in a special way. This often is in contradiction with the actions which are necessary to optimize database performance. In order to overcome the data modeling problems which are recognized at time of database maintenance, we developed an approach to database optimization at the conceptual level. We use an extension of the entity-relationship model, the Higher-order Entity-Relationship Model (HERM), and the workbench RADD developed for supporting HERM specifications. Although a high level of abstraction is provided in order to be user-friendly, in RADD data structures and integrity constraints can be specified together with data profiles, operations, and other application requirements. In the thesis we present and discuss the RADD/raddstar system, which is the subsystem of RADD used to specify additional processing requirements, to evaluate and verify behavior properties, and to optimize the conceptual schema. We invest special interest on previous approaches to database modeling and database optimization, and define the data model that is used for RADD/raddstar’s internal evaluations (RADD*). RADD* represents the items of the conceptual and internal schema by functional terms, and enables the user to add database application functions and behavior specifications to the graphical RADD database design. This way, we are able to analyse maintenance aspects of the designed schema and to find possible contradictions and performance bottlenecks. With or without the additional requirements that are specified by the database designer, the generation of schemata for implementation under consideration of, but independent from a specific DBMS can be used for the analysis of the given design, for the discussion of bottlenecks, and for the generation of design schemata with better operational behavior.