Data Warehousing is a concept that grew out of the need to store and analyze large amounts of data over long periods
of time. Most software applications are designed to store a limited amount of data, requiring historical information to be
purged periodically. Most software also does not store the various versions of a data element as it changes (is updated)
over time. For these reasons a great deal of analytical value is lost if a data warehouse is not developed and
implemented. Organizations seeking to become more strategic in their planning and execution are compelled to implement
some form of data warehousing.
Data warehouses can take a variety of forms and there are several leading methodologies for how data warehouses should
be designed and implemented. The most common forms of data warehouses include the following:
-
Data Mart: A special database system design that lends itself to slice and dice operations. Data marts are typically
developed to address the analysis needs of a specific subject area within the organization. More recent versions of this
concept allow for a more comprehensive view by incorporating several subject areas into a single dimensional database
design.
-
Operational Data Store (ODS): A temporary storage area for data from a variety of operational systems such as ERP, CRM
and others. An ODS typically represents integrated data across subject areas and can contain conventionally designed
relational database designs as well as dimensional designs for analytical applications. This sometimes is used as a
"stepping stone" towards a full data warehouse approach because an enterprise data warehouse (EDW) can be updated from
an ODS periodically.
-
Enterprise Data Warehouse (EDW): A comprehensive database system that stores data in perpetuity and provides a single
unified source for all types of data analysis, reporting and visualization. EDW's often include time-varied data which
represents the changing status of individual transaction records. An EDW usually feeds data marts and other analytical
sources, but can be accessed directly as well. In some cases reference data that is not found in other systems and is
sometimes referred to as Master Data is also stored in an EDW.