Data Vault and the Oracle Reference Architecture
Thanks to Mark Rittman and Twitter, I found out just before RMOUG that Oracle had published a new reference architecture. It used to be called the Data Warehouse Reference Architecture, now it is called the Information Management Reference Architecture.
Oracle updated the architecture to allow for unstructured and big data to fit into the picture.
In my talks about Data Vault over the last few years I have been referring to the Foundation Layer of the architecture as the place where Data Vault fits. The new version of the architecture actual fits the definition of the Data Vault even better.
Now the Foundation Layer is defined as “Immutable Enterprise Data with Full History”.
If that is not the definition of Data Vault, I don’t know what is!
Immutable - does not change. Data Vault is insert only, no update – ever.
Enterprise Data – well duh! That pretty well fits any real data warehouse architecture. The model covers an enterprise view of the data not just a departmental views (like a data mart).
Full History – tracks data changes over time. That is one of the keys to the data Vault approach. We track all the data change history in Satellites so we can always refer to a view of the data at any point in time That allows us to build (or re-build) dependent data marts whenever we need or whenever the business changes the rules.
So it is possible to do a Data Vault approach and be compliant with Oracle’s reference architecture.
Guess Dan was just a bit ahead of the game…