What is the fundamental difference between the approaches of normalized storage (according to Inmon) and multidimensional storage (according to Kimball)?

Question

Database

Good afternoon!
What is the fundamental difference between the approaches of normalized storage (according to Inmon) and multidimensional storage (according to Kimball), if:
when using the approach (Kimball) and the snowflake schema, dimension tables are still normalized?
I would be grateful for examples or materials for study.

Answer 1

The fundamental difference in the speed of development and changes.
Let's say we need to give several showcases about sales with slightly different logic to different customers.
Kimball: Let's build them on the springs! Yes, they will have very similar loading logic, but we will do it quickly. Alas, in this case we most likely will not succeed in a single truth. When the storage becomes large, you will have to spend a lot of effort on support. changes to the source will need to be accounted for in several places.
Inmon: First, let's normalize everything and make a single version of the truth. In this case, we will expand all sales in a single form for all. and above it we will already build all the showcases that users need, where we implement exactly the logic that they need.
In fact, you can start building a repository based on any approach. If you start with Kimball, then sooner or later some intermediate tables with a single truth will be born and the model will become similar to a datavault. If you start with a datavault, then you still sometimes have to make custom storefronts on sources when you need not a single truth, but exactly the logic that the user wants. Thus, in reality, we get a mixture of these two approaches.