Enterprise Data Modeling Practices - High-Performance Web Databases: Design, Development, and Deployment - page 161

Data Administration as a Service Provider
A Data administration group can be of enormous use to project teams, or
it can be an enormous source of frustration. If a Data administration group
comes up with some sound solutions or possible designs before a project
starts with specification of details, it can be of great assistance. The Data
administration department is a service and not a control function. Manage-
ment may want to ask:
• Is the Data administration a service provider or only a review institution?
• Do they have a written statement that they are a service provider?
• What do the project managers think about the service provided by
Data administration?
• What is the ratio between services and control functions in the Data
administration department?
Integration of Data Modeling Activities into the Process of Specification
Data modeling should be an integral part of the specification process
right from the beginning. This practice will promote reuse and avoid revis-
iting old arguments. To ensure that data modeling is properly integrated,
managers should ask themselves:
• When do the project teams usually meet the Data Administration
group — at the start of a project or at the end in the review?
• Are there separated project data models?
• Are there people who do data modeling on a C level for their own sake
without a project needing it?
• Is there a mechanism that allows projects to borrow parts of the data
model for rework and new developments?
• How do projects integrate their results into the global master enter-
prise data model? Is there any assistance or tool for integration?
• Who supervises the process of integration?
Separated Levels of a Data Model
Reference models separate clearly between different levels of abstraction
in a data model. This is required because different levels are different blue-
prints for different groups of people (A, B levels for business administra-
tion, C for data systems professionals). The Data administration group
should ask:
• Do we have level A and B data models that are more than just a pure
copy of a reference model?
• What do the users think about data modeling?
• Is there a repository that strictly separates levels of the data model?
• Will users working on projects see at once which level they are cur-
rently looking at when dealing with the dictionary?