From a Business Perspective
- A business is a set of business processes
- Business processes can be either
- Operation — “Make it work”
- Transactional — “What is going on?”
From a Technical Perspective
Kimball’s Definition
A copy of transaction data specifically structured for query and analysis
Inmon’s Definition
- Subject-oriented — May need different DWH for different subjects
- Integrated — Comes from potentially multiple data sources
- Time-variant — The data changes with time
Rainardi’s Definition
- Updated periodically in batches into a dimensional or normalized form
DWH Architectures
Kimball’s Bus Architecture
- Broken down into the back room (kitchen) and front room (dining area)
- Top down approach where back room decides what front room receives
Independent Data Marts
- All departments operate independently
- Inconsistent ETL may lead to inconsistent data conclusions
- Generally discouraged
Inmon’s Corporate Information Factory (CIF)
- A better kind of data mart where data marts can get whatever data they want, from a standard source of truth situated in the EDW
- Uses 3NF for back room
- Data marts can access the back room for other data fields if needed
Hybrid Bus & CIF
- Still uses conformed dimensions, like the bus architecture
- However, bus can access back end where the 3NF EDW sits
