Archives of the TeradataForum
Message Posted: Tue, 20 May 2003 @ 12:12:02 GMT
This answers your first query...
Normalization has a cost. It optimizes update performance at the expense of retrieval performance. But a normalized schema is very flexible & can answer questions not previously asked or thought of & needless to mention it eliminates all the anomalies. As a result of normalization, the relations will usually be smaller & joins are frequently required to satisfy queries. Teradata, by virtue of its parallel architecture can handle joins easily without major performance hits but performing joins is one of the main pain areas of other RDBMS products which is why they advocate denormalized schema like Star or Snowflake.
A warehouse ideally should be in a position to answer "adhoc" queries & a 3NF schema is in a better position to do this as compared to "Star Schema".
Star schema is ideal when the questions asked of the database are all know beforehand.
For more info, refer Teradata Design (Chap. 5).
|Copyright 2016 - All Rights Reserved|
|Last Modified: 27 Dec 2016|