Reference models for enterprise-wide data engineering

               Hars, A.; Langemeyer, J.; Scheer, A.-W.
                 Institut fuer Wirtschaftsinformatik
                 Universitaet Saarbruecken, Germany
                    e-mail: hars@iwi.uni-sb.de
                Phone: +49 681 302-4297 Fax: -3696

Abstract

This paper outlines a model-based, unified approach for enterprise
integration focusing on the data view which is currently being
developed by the European project CODE (Computer Supported
Enterprise-wide Data Engineering). The approach proposed extends
the existing data modelling methodologies to support the
utilization of predefined reference data models. This implies the
design of a library for reference data models, the specification
of a customizing process that allows to derive particular data
models from the reference models stored in the library and the
development of the corresponding the tools and computer-supported
methods. The reference model approach is limited to the data view.
It is expected, however, that important parts of the methodology
for the customizing of the data models shall also be applicable to
function, process and organization models.

The paper shall briefly outline the project's background: an
enterprise-wide data model for a typical industrial enterprise
developed by SCHEER1, which consists of approximately 300 entity
types and relationship types, is used as a first approximation of
a reference model. It is structured into modules which can be
recombined to form the original enterprise-wide data model. This
enterprise-wide data model has been used in several industrial
projects.

In a second step the customizing process is outlined and an
example given. The library of reference models is the starting
point for the creation of enterprise-wide data models. This can
already be accomplished by using existing data dictionary systems.
The benefits of reference models can only be realized, however, if
interactive procedures are available, which support the process of
adapting a reference model to the specific situation of a certain
enterprise. The amount of information to be treated during this
task, the frequent variations of the models to be created, complex
consistency checks and the need for visualization require computer
support. This process is termed customizing. It consists of a
series of steps which are subsequently described:
1. Enterprise description: This step prepares the selection of
area reference models by describing the enterprise according to a
predefined typology.
2. Selection of area reference model: From the description of the
enterprise, a set of possible area models which might be
appropriate is derived and presented to the user. He analyses the
different models, compares them with the enterprise situation and
finally decides for the most relevant area models.

3. Iterative adaption of the area reference model: The adaptionprocess consists of three parts: formal changes as consequences of
adaptions which are proposed by the computer, adding structures
which have not been part of the reference model, but are relevant
for this enterprise and deleting of structures from the reference
model which are not relevant for this enterprise.
4. Evaluation of area model created: Each area model has to be
checked for consistency. This comprises syntactical correctness
and might also include a cross-check by a third person.
5. Merging of area models to enterprise model: All customized
area-specific models are merged to become an enterprise-wide data
model.
6. Validation of enterprise-wide data model: In a final step, the
model generated has to be tested. Checks are oriented towards
completeness and uniform level of detail.
7. Transformation to logical/physical structures: This process is
one of the many uses of the customized enterprise-wide data model.
It allows the transformation into data structures conforming to
one of the logical data models hierarchical, network-oriented or
relational model and subsequently the transformation into the data
description language of a specific database management system.
This functionality already is common in many tools.

In a final section, the key assumptions of our approach are
described and an attempt is made to relate our approach to the
architecture proposed by Cavalli and Smith2. This leads to a short
critique of the architecture.


____________________
1 Scheer, A.-W.: Enterprise-Wide Data Modelling - Information
Systems in Industry. Berlin 1989.
2 Cavalli, R.; Smith, R.: Technical Issues of Enterprise
Integration CAD-349-91. Report 1991.