Cloud data warehouse architectures

Kimball's Bus Architecture

• Results in a common dimension data model, that is shared by different departments.
• The data is not kept at the aggregated level, but rather at the atomic level
• Organised by business processes and used by different departments

Independent Data Marts

• Departments have independent ETL processes and dimensional models, called "Data Marts".
• Different fact tables for the same events, no conformed dimensions.
• Uncoordinated efforts can lead to inconsistent views.
• ETL processes designed by specific business departments for their needs. This might by faster, but with the downside of inconsistency

As independent data marts are uncoordinated and inconsistent it is not encouraged to use them.

CIF - Inmon's Corporate Information Factory

• Data marts have the same baseline.
• Data marts are dimensionally modelled, mostly aggregated
• The users can get data from the data marts or the 3rd Normalform database
• There are 2 ETL process done -- source systems -> 3 Normalform database -- 3 Normalform database -> Department data marts
• The 3 NF DB is the enterprise wide data store

Hybrid Bus & CIF

• uses the 3 Normalform database like in CIF
• uses conform dimensions like in Kimball's bus architecture