Enterprise Data Modeling Practices - High-Performance Web Databases: Design, Development, and Deployment - page 158

• Cannot provide a bank with information systems that adapt with the
speed of innovation. New financial instruments require fast adaptation
of IS capabilities.
• Acts as a brake and not a booster for rapid systems development.
• Promotes structures that should not be changed. Change is the norm
in systems development. This adds to the characterization of data
modeling as a brake.
• Creates additional complexity in the process of software develop-
ment. Applications development is impaired by the use of top-down
data modeling practices.
• Can lead to the violation of normalization rules, with the consequent
need to adapt the integrated systems.
• Has to incorporate all possible future options of a system, which slows
operation down to a standstill. The level of abstraction tends to in-
crease indefinitely until no one is able to understand the results of the
process.
• Is useless, if purchased off the shelf as prefabricated data models.

STARTING POINTS FOR BETTER DATA MODELING

The following four requirements should be the starting points for data
modeling. These goals are essential to a reasonable process of software
development. Despite the criticism, these basic requirements are often
acknowledged as essentials of any software development process:

• Integrated systems are a prerequisite for the survival of a financial in-
stitution, to manage complexity and to master interdependencies.
Data modeling was invented to integrate systems and important termi-
nology. The goals of this effort are reuse and data integration.
• The separate systems of an enterprise must use consistent terms to
provide a consistent processing of data across the boundaries of sev-
eral systems.
• Integration of old and new systems is necessary for routine systems
development. As system integration is the norm in systems development,
bottom-up strategies and reengineering of old systems must be support-
ed. Good data modeling practice will provide support for this process.
• Fundamental structures or invariants of a business are the basis for all
systems development.

Is Enterprise Data Modeling Really That Bad?

The rebuttals to the arguments will show that most of the problems with
data modeling can be fixed by a data modeling process that is oriented
toward goals.

Rebuttal: No attempt
should be made to model transitory facts in a data model. Hardwired

Data Modeling Does Not Accommodate Rapid Change.