Computer-assisted logical database design

Abstract

Database design can be a long and tedious process. Designing large databases, say of 500 elements or more, usually takes months, even years. Until now, database design has been more an art than a science, with no standardly recognized set of rules or algorithms. Experience and intuitive feel have been the designer's criteria. In addition to the time and tedium involved, the quality of the resulting design is frequently a problem. In many cases fo poorly performing systems, the database design is found to be the culprit. Once the meanings of the data elements are understood, database design is relatively straightforward but tedious. But good designs are not always easy to obtain. The major problems fall into the following categories: • • insufficient analysis of functional data requirements • • lengthy tedium beyond human capacity • • difficulty of predicting performance implications • • management impose time constraints Automated logical database design offers the potential of quality design accomplished in significantly shorter periods of time. Besides assisting in the structuring process, the procedures offer a wealth of editing and diagnostic information. While automated techniques can derive a logical datastructure, the designer should always consider these techniques as a tool that produces a suggested (cannonical) design requiring human review and probably human refinement. These techniques, then, should be applied to an interative design process to assist the designer in converging on a final logical design to be implemented. The available diagnostics can greatly assist in this review and refinement. The emphasis of this paper is on the design of DL/I hierachical structures, but the design of relational structures is also treated. Following some introductory concepts, the automated logical design procedures for DL/I structures are presented. The procedures are described, some of their more subtle features are explored, and the types of reports obtainable are suggested. The extension of these concepts to relational structures is then presented. Finally, desirable interfaces between the automated design procedures and a dictionary system are suggested.