Enterprise Data Modeling Practices - High-Performance Web Databases: Design, Development, and Deployment - page 165

of access layers is not possible without a rigid data modeling practice. The
effort saved by not performing access layer programming is a very tangible
effect of data modeling. It should be determined whether the software
development department uses access layers, and if so, whether they are gen-
erated or at least standardized.
Data Types
Most COBOL or PL/I shops do not have a concept of data types. This
leads to enumeration types being redundantly described in textual
attribute descriptions, structures being described as entities, and other
design flaws. It is helpful if an organization has a data type concept that is
supported by code-generation facilities and a repository that allows a data
type concept.
Data Dictionary Product Properties
It is important for data dictionary software to support requirements
concerning the separation of model layers, version concepts, and data
types. It is a good sign if the software developers and project teams like the
look and feel of the data dictionary. A data dictionary should:
• Support data types such as enumeration types, records, range types,
and a code generation facility.
• Be adaptable to changes in its metamodel. Improvement of the soft-
ware development process can be slowed or made impossible by an
inflexible dictionary.
• Have a version concept.
• Support integration of project data models. This is best done using
check tools that support the Data administration group.
• Support separation of levels, documentation of transformations between
levels, and separated production and reference models.
CONCLUSION
A broad range of criticism has been brought against data modeling. This
article focused on the criticism to derive several lists of critical questions
concerning organizational, technical, and quality success factors for a
good data modeling practice. These questions were used to check the data
modeling practice of a large financial institution. Exploring the issues cov-
ered and considering the questions posed should help Data Administration
groups avoid some of the problems common to data modeling.