Denormalisation is the
keystone of
data warehousing, in which data from one or more
tables of one or more
databases is denormalised and stored in a large
warehouse database.
Because the denormalised tables represent the result of table joins commonly performed during operational
application - but which may be slow due to their iterative nature - data warehouse databases tend to be generated periodically (nightly, weekly).
An easy way to
imagine a data warehouse is to think of all the joined queries you would perform on your database, then create tables that represent those queries. Because no
join is necessary, the warehouse database is much faster at returning result sets.