Data mart

from Wikipedia, the free encyclopedia

A data mart is a copy of the partial data stock of a data warehouse (DW) that is created for a specific organizational area or a specific application or analysis (see below). It can also be understood as a partial view of the data warehouse or a non- persistent cache. In practice, in some cases the data stock in a data mart is also kept for the long term (see also "independent data marts").

Reasons for working with a data mart ( "copies" from the data warehouse) instead of direct access to the data in the data warehouse based on views ( views ) are:

  • the need for special data structures (which are not available in this form in the DW) for certain analyzes, e.g. B. for multi-dimensional analysis, so-called Online Analytical Processing (OLAP)
  • better performance ( Performance ): shift of computing power to another computer and / or relocation of accesses to a different memory and / or in the case of local use less network load
  • Independence of the user (e.g. mobility, independence from other organizational areas)
  • Access protection : delimitation from other users or opening up to other users (the data in the data warehouse is not accessed directly, but only indirectly via the data mart)

By using data marts, the respective application area (technical), the target system (technical) and the user (organizational) can be largely taken into account.

The data stock of the data warehouse is distributed through the data marts. The data marts created in this way require the data warehouse, which is why they are referred to as dependent data marts.

In addition to this data mart concept, there is also the historically relevant concept of the independent data mart, which exists separately from a data warehouse and which enables a view of one or more source systems. This architecture approach runs counter to the idea of ​​data warehousing because it can lead to many isolated solutions over time. It comes from a time when the central data warehouse, the data warehouse, did not yet exist.

The data model of a data mart often corresponds to a multidimensional or OLAP structure. The technical implementation of this data model can take place as a multidimensional database , as a relational database or, increasingly, in the main memory of the computer.

See also

literature

  • A. Bauer, H. Günzel: Data warehouse systems - architecture, development, application. dpunkt, 2013, ISBN 978-3-89864-785-4 .
  • W. Lehner: Database technology for data warehouse systems, concepts and methods. dpunkt, 2002, ISBN 3-89864-177-5 .