Data Mart/Dimensional Data Store Definition.

[MarkW]

Hi,

My understanding of a dimensional data store is one where it contains a number of data marts, made up of fact and dimension tables only. Is this correct? Is it possible to have a data mart that does not consist of fact and dimension tables? Is it possible to have a dimensional data store that's not made up of data marts?

Funny question I know, but looking for a bit of a sanity check.

Thanks,

M.

[ngalemmo]

Dimensional modeling is a methodology to achieve a data model. Entity relationship modeling is also a methodology to achieve a data model. They are different methods that result in different models of the same thing. Methodologies exist for a reason. You can, of course, use any method you want to create any collection of columns and tables you want. Some are more effective than others.

[MarkW]

So, would a data model based on entity relationship modelling suffice to call something a 'data mart', or better put a collection of 'data marts', and therefore, the sum of all those parts a 'dimensional data store'. Surely not?

Regards,

M.

[ngalemmo]

Given 'data mart' is an incredibly vague term, any table or tables can be considered as such. Different architectures use the term differently.

In Inmon's architecture, he breaks things down into a somewhat normalized enterprise DW and 'marts'. To him, a mart can be anything from a star schema, a normalized structure, to a flat table. A mart is simply something consumed for analytics.

In Kimball's architecture the term 'mart' refers to a single star schema (a single fact table and related dimensions). A dimensional data warehouse is a collection of such marts.

Lindstedt's data vault is a store and publish architecture like Inmon's, he as a similar vague definition of 'mart'.

[Sponsored content]