difference between data mart and data warehouse at logical/physical level

hi at all

I'm a bit confused about the difference between data warehouse and data mart.

From what it reads around internet and books the differences between data mart and data warehouse are about the follows:

-) A data warehouse is created at the enterprise level (organization level). A datamart is created at the department level/business division.

-) A datamart is a subset of the data warehouse and are built for specific user group.

But, what i want to know, if possible, is: at the logical/physical level, what is the difference between data mart and data warehouse?

What I know is that the logical model of data mart is the Star Schema (that is a dimensional model). And for the data warehouse, what's the logical schema/model?

thanks.

You have every right to be confused. There is no one clear definition of what a data mart is. It is essentially a useless term unless placed in specific context.

If you limit the definition as used by Ralph in the first edition of ToolKit, a mart is a single star schema. That is, a fact table and its related dimensions. A data warehouse is a collection of data marts (a foundation concept of dimensional design).

There are dozens of other interpretations of what a data mart is.

ngalemmo wrote:[...] If you limit the definition as used by Ralph in the first edition of ToolKit, a mart is a single star schema. That is, a fact table and its related dimensions. A data warehouse is a collection of data marts (a foundation concept of dimensional design). [...]

So, in this case, is a data warehouse a collection of related star schemas at logical level?

If yes, how are they related? ... I imagine they are related with foreign key between fact tables. it's right?

Sorry, but i need to understand this concept in depth.

No, they are related by conforming dimensions. You do not set up FK's between fact tables. Any relationship between facts is considered many-to-many. BI tools that support dimensional models recognize this and combine facts by first summarizing each fact along common dimensions (those used in the query) before combining the two (either by a join or union).

Ok, thanks for your explanation.

Now, it's a little more clear.