New features for a relational database system to support computer aided design
Antonin Guttman · 1984
Data management is a significant problem in Computer Aided Design (CAD) systems which handle large volumes of data in a variety of forms. However traditional database systems have not been widely used in CAD because they are hard to use and do not offer some needed functions. The purpose of this research is to develop features that can be added to relational database systems to make them more useful for supporting CAD. We ran an experiment in which a prototype relational database system, INGRES, was used to store Very Large Scale Integrated circuit (VLSI) design data, and to perform typical CAD retrieval operations. INGRES was awkward to use in this context and ran much slower than specially designed CAD programs. The remainder of this thesis investigates several additions to INGRES suggested by the experiment that would make it easier to use and more efficient. We augmented the relational query language QUEL with an append* command for expressing transitive closure queries, which are needed to retrieve hierarchically structured VLSI designs. Processing issues are discussed and implementation test results are given. An indexing structure called an R-tree is presented which can be used to retrieve data for overlapping non-point spatial objects according to their locations in a multi-dimensional space. R-trees are designed for data residing on paged secondary memory, and are suitable for use as a database access method. Specialized applications like CAD often assign meaning to data that cannot be exploited in a general-purpose database system to optimize data structures and query processing. We outline a way for user-supplied semantic routines to supply information to the database system for this purpose. CAD applications often use composite design entities made up of many parts. We discuss three ways to represent such composites in a relational database: as abstract data type arrays, by means of built-in array types, and as nested relations. All the features we have studied could be added to current relational database systems, and would be useful not only for CAD but also for other applications involving spatial and highly structured data.