Comments on: Data Modeling for Analytical Data Warehouses. Interview with Michael Blaha. Trends and Information on Big Data, New Data Management Technologies, Data Science and Innovation. Mon, 03 Nov 2014 17:39:20 +0000 hourly 1 By: Reinhold Thurner Wed, 07 Mar 2012 15:40:56 +0000 Michael writes “Many data warehouses contain sensitive data..” . The same is true of course also for operational data.
COBIT ( addresses the issues of compliance and security related to these data with the process P02 “define information architecture”. This process “creates and regularly updates a business information model; This encompasses the development of a corporate data dictionary” etc. The obligation of IT on corporate level is to maintain proper records about the “data syntax rules, data classification scheme and security levels” (beyond modeling on the project level). COBIT seems to have in mind some sort of global “conceptual model” including information about the physical instances and the access rights.
The “Common Warehouse Metamodel” (OMG) on the other hand is a metamodel for the description of the physical instances of operational data, the warehouse and the mapping between both. The CMW relies (no surprise) on MOF to describe the metamodel. The final outcome is – in contrast to the “data dictionary view” of COBIT – a recipee for the creation of mapping services. This however would not be sufficient to support the reporting and auditing requirements of the IT-Auditors from ISACA. It would be necessary to combine both – i.e. a CWM-Dictionary which contains metadata also for auditing and controlling purposes. Such a dictionary could obviously be the basis for specs or even model driven generation of mapping services.
What is the main problem? Michael said “Data warehouse metadata includes source-to-target mappings, definitions (of facts, dimensions, and attributes)”. I think the “relationships” are the biggest challenge. They are the real source of the complexity of this task: E.g. the relationship between an attribute of a (conceptual) entity and the implementations of this attribute by various databases / columns.
We speak often of the so called “impedance mismatch”. There is such a mismatch between the corporate data governance view (COBIT), the MOF-View of the CWM (still quite global) and the views of individual projects / applications (which deal with the real world details). A “real corporate dictionary” could brigde this gap – but this is quite a challenge!

By: Rick Tue, 06 Mar 2012 17:50:27 +0000 I still have “Object-Oriented Modeling and Design” on my bookshelf.

Thanks Michael.