Design Considerations
Whether third party vendors from whom you are considering buying tools are Teradata
partners. Query tools, for example, should take advantage of the Teradata Database
optimizations (for example, they should incorporate the standard Teradata Database
aggregate and ordered analytical functions rather than performing in-line calculations
of OLAP statistics).
If you intend to perform a mix of ad hoc tactical and decision support queries or
if you plan to undertake any serious data mining projects, then you should limit the
physical denormalization of the database to a minimum.
For example, if you perform frequent ROLAP analyses of your data and performance is
not what you expect to see, consider off loading the data to a dependent data mart
designed around the multidimensional model that many OLAP proponents advocate. You
should not denormalize the entire database just to support a few OLAP applications.
If you plan to use data mining technology, then a normalized database is the key to
your success. Data mining techniques do not perform well in a denormalized environment.