Data Mart vs Data Warehouse
Some time ago, I started to understand OLAP and in this post I want to check the correctness of my own thoughts on the basis of these two concepts.
Terminology
Data Mart is a data showcase in translation.
Data Warehouse - data warehouse.
But the fact is that both of these terms can be translated as
Data Warehouse ... Data Mart - a slice of Data Warehouse. (clickable) There are two approaches to building data warehouses: The first is simpler, according to Kimball (Kimball) Here the information from the OLTP system directly goes to the data mart, from where OLAP takes the data it needs. The first case obviously illustrates the approach of M. Demarest (M.Demarest), which, according to Wikipedia:
Second, according to Inman, see book.
Here, information from OLTP first goes to the data warehouse, then only to data mart, from where OLAP again takes the information it needs.
We are adapting the Inman approach, which calls the data warehouse a real religious database, while Kimbel is a mixture of data marts.
Data mart is a database created according to the requirements of dimensional modeling and consists of fact tables and measurement tables.
Data Mart
When implementing the system according to the Kimbel methodology, the frontend of the database should be data mart, which uses Analysis Services for the cube as a
Data Warehouse data source (clickable)
thus, it is more complicated than data mart, it includes not only a database, but also a decision support system and a client-server architecture, whereas data mart is essentially a database created taking into account the requirements of future cubes.
Sources:
http://ru.wikipedia.org/wiki/Vitrina_dannyh
http://ru.wikipedia.org/wiki/Hranilische_Dannyh
Book Expert Cube Development with Microsoft SQL Server 2008 Analysis Services, in English which I translate in his blog.
Terminology
Data Mart is a data showcase in translation.
Data Warehouse - data warehouse.
But the fact is that both of these terms can be translated as
Data Warehouse ... Data Mart - a slice of Data Warehouse. (clickable) There are two approaches to building data warehouses: The first is simpler, according to Kimball (Kimball) Here the information from the OLTP system directly goes to the data mart, from where OLAP takes the data it needs. The first case obviously illustrates the approach of M. Demarest (M.Demarest), which, according to Wikipedia:
in 1994, proposed to combine the two concepts and use the data warehouse as a single integrated data source for data marts.
Second, according to Inman, see book.
Here, information from OLTP first goes to the data warehouse, then only to data mart, from where OLAP again takes the information it needs.
We are adapting the Inman approach, which calls the data warehouse a real religious database, while Kimbel is a mixture of data marts.
Data mart is a database created according to the requirements of dimensional modeling and consists of fact tables and measurement tables.
Data Mart
When implementing the system according to the Kimbel methodology, the frontend of the database should be data mart, which uses Analysis Services for the cube as a
Data Warehouse data source (clickable)
thus, it is more complicated than data mart, it includes not only a database, but also a decision support system and a client-server architecture, whereas data mart is essentially a database created taking into account the requirements of future cubes.
Sources:
http://ru.wikipedia.org/wiki/Vitrina_dannyh
http://ru.wikipedia.org/wiki/Hranilische_Dannyh
Book Expert Cube Development with Microsoft SQL Server 2008 Analysis Services, in English which I translate in his blog.