Although not required, it can be useful to build a database strictly for analytical queries, separate from the transactional database. When a data warehouse is loaded, information from operations on a transactional database are transferred, often millions of records at a time. This volume of data overloads traditional normalized databases. Additionally, data warehouses generally don't need the added overhead of transaction logs, rollback logs, and incremental referential integrity checking inherent to transactional databases. The basic steps for the creation of a data warehouse are:
Plan a design.
Create tables.
Move data to the warehouse.