The rule of thumb in data architecture is that “"if you could have one, do not build two.” However, what are the criteria to determine if an organization only needs one warehouse?

PCWeek Magazine July 6, 1998, Volume 15, Number 27 by Jeff Mead

Many large corporations that have many lines of business are decentralized. Within the domain of a line of business, its enterprise data warehouse contains only the data pertaining to this line of business. From the standpoint of the entire corporation, this LOB data warehouse is just a repository of one subject area.

It is not always practical to build just one enterprise data warehouse in many cases. How do we ensure that each LOB data warehouse is able to cross reference others in the entire enterprise? Federated data warehouse is the answer. A federated data warehouse is

Promotes information sharing across Our Company enterprise

To build a federated data warehouse environment, the organization must address the following issues:

Data Consistency
- Data must be in a relational structure
- Meta data must be maintained in corporate repository
- Shared data must conform to enterprise standards (for example, naming, codification, common dimensions, etc.)

Technology Consistency
- Hardware & software platforms must be compatible
- Tools must follow company standards
- “Lowest common denominator” functions of database management systems must meet minimal requirements across the enterprise (vs. departmental requirements)
Process Consistency

- Warehouse/mart management and other related administrative activities must share the same process rules and support infrastructures. This includes ETL (extraction/ transformation/ loading) activities, warehouse administration, production control, etc.