Data Vault Modeling
Today the Data Warehouse needs to be Agile. While there are many barriers to achieving Data Warehouse program agility, one of the issues has been the Data Modeling approach. With traditional techniques, the Data Warehouse quickly becomes hardened and difficult to change. Moving to Agile development techniques, based on incremental builds, is almost impossible. Why? Because we ultimately need to do a great deal of re-engineering of the data structures.
Enter Data Vault (DV). For the past 15 years companies around the world have been using a new Data Modeling technique that greatly improves agility - the Data Vault data modeling approach. The premise behind Data Vault is Unified Decomposition - basically this means we separate the things that change from the things that don’t change. How this works: The existence of a person named “Hans” (for example) is always going to be true. So, the instances of Core Business Concepts (CBCs) are placed in their own data structures (Customer Hub, for example). Next the innate relationship that the Customer has with a Sale is also something that is not generally subject to change. This is a business-driven, foundational relationship that we capture in unique table structures (Links). Since no relationships are embedded they can be added without any re-engineering impact. Lastly the way we describe our CBCs can vary over time, vary by source, vary by type of data and also by rate of change. We use a set of separate tables to capture this context (Satellites). Because new attributes introduced in later iterations can be included in new Satellites, the Data Warehouse can accept new attributes without re-engineering. There are over 1500 Data Vault models in organizations today and the technique is growing rapidly.
The Data Vault modeling pattern is also very applicable to Big Data, Cloud, Virtual and Streaming deployments. Because the context is separated from the way CBCs and Relationships are stored, the context can take any form.
What you will learn
- The foundational pillars of the Data Vault modeling approach
- Through small group modeling exercises learn to translate requirements to DV models
- How to model a Data Warehouse using Data Vault (DV)
- To distinguish between encapsulated (3NF & Dimensional) versus ensemble (DV, etc.) modeling patterns
- To identify modeling scenarios that are best addressed by Data Vault
- To review and critique DV models for best practice compliance & optimal performance