Microsoft Office Tutorials and References
In Depth Information
Figure 1-10. Data warehouse and data marts
A data mart is a baby version of the data warehouse. It also has cubes embedded in it, but you can think of a data mart
as a store on Main Street and a data warehouse as one of those huge, big-box shopping warehouses. Information from
the data mart is consolidated and aggregated into the data-warehouse database. You have to regularly merge data
from OLTP databases into your data warehouse on a schedule that meets your organization’s needs. This data is then
extracted and sent to the data marts, which are designed to perform specific functions.
■ Data marts can run independently and need not be a part of a data warehouse. they can be designed to
function as autonomous structures.
Consolidating data from a data mart into a data warehouse needs to be performed with utmost care. Consider
a situation where you have multiple data marts following different data schemas and you’re trying to merge
information into one data warehouse. It’s easy to imagine how data could be improperly integrated, which would
become a concern for anyone who wanted to run analysis on this data. This creates the need to use conformed
dimensions (refer to http://data-warehouses.net/glossary/conformeddimensions.html for more details.) As
we mentioned earlier, areas or segments where you map the schemas and cleanse the data are sometimes known
as staging environments . These are platforms where you can check consistency and perform data-type mapping,
cleaning, and of course loading the data from the data sources. There could definitely be transactional information in
each of the data marts. Again, you need to properly clean the data and identify only the needed information to migrate
from these data marts to the data source.
Search JabSto ::