Data is extracted from the data sources (DBs, files, APIs, …) and stored into a staging area for further processing, cleaning and transformation. Next, the data is loaded into the data warehouse, together with its metadata (origin, timestamps, …). Summary data is preaggregated data that allows faster analytics. Different users access the same data via different data marts. A data mart contains a subset of the data warehouse for specific needs.
Variants
Independent Data Marts
- Data is directly loaded from the sources into the data marts
- Multiple “mini data warehouses”
- No single point of truth
Virtual data marts
- Data marts as (logical) views on the data warehouse