[Back to Papers]

Efficient Database Implementation of EXPRESS Information Models


Efficient Database Implementation of EXPRESS Information Models
PhD Thesis, David Loffredo

I have made PDF versions of my PhD thesis and defense slides available for download. Although I am happy to have you download, read, and use this, please note that re-distribution of this document is not allowed without permission. If someone else would like a copy, please send them back to this page.

If you plan to reference any of this material in another document, the proper citation is:

Loffredo, David, Efficient Database Implementation of EXPRESS Information Models. PhD Thesis, Rensselaer Polytechnic Institute, Troy, New York, May 1998.


The research presented in this thesis describes a framework for database implementation of EXPRESS information models. EXPRESS models describe complex structures and correctness conditions for engineering activities, and are defined by the ISO-10303 Standard for Product Data Exchange (STEP). These models are a substantial engineering resource, and industry desires to use them to integrate design and manufacturing processes. Databases built around STEP models are essential because they provide content that integrated engineering processes understand.

The Standard Data Access Interface (SDAI) is a STEP API for EXPRESS-defined data. Prototypes have attempted to provide SDAI access by implementing each SDAI operation as one or more native operations directly upon the database. A direct binding can be costly, as it requires completely new software for each database. This work proposes several SDAI implementation architectures that offer alternatives to a direct binding.

To evaluate the real-world performance of implementations, this work defines a set of representative benchmarks on the STEP AP-203 information model. AP-203 contains information such as CAD geometry and product configuration that is common to all of the STEP models. The STEPStone benchmarks cover information that is modeled in an existence dependant style (PartStone, part versions), a navigational style (NURBStone, geometry), and a mix of the two (BOMStone, bill of material).

The results of timing experiments using these benchmarks are presented. The experiments evaluate the performance of direct-binding SDAI implementations built on relational and object-oriented databases, and examine the effect of various optimizations on binding performance. Analysis of the timing results provide the relative cost of access for each system, and allow us to determine when each implementation style will be most advantageous. In addition, these experiments provide insight about the use of SDAI access versus traditional access strategies.