Database design by computer-aided schema transformations

Abstract

The focus of the paper is schema transformation during the development of an information system. A framework is described for conversion and transformation of conceptual (semantic) data models and their internal (machine-oriented) representations. This framework allows us to ‘walk’ through the solution space of candidate internal representations for a given conceptual data model. This walk may be randomised or performance-driven, where storage requirements and average response times are combined in a multi-objective fitness function. Furthermore, a wide variety of control parameters may be embedded, such as preferences for database table size, absence of data redundancy or absence of optional database fields. Basic experimental results produced by a prototype convertor/transformer are presented, including deviations from the standard optimal normal form for databases.