Archives of the TeradataForum
Message Posted: Fri, 08 Nov 2002 @ 16:41:41 GMT
| Subj: || || Re: 3NF vs Dimensional modelling |
| From: || || Myers, Nancy |
The short answer is: 3NF is preferable for data warehousing if the platform will support it performance-wise.
Dimensional modeling requires that raw data be distilled/summarized, thereby losing detail along the way. This runs contrary to the data
warehouse goal of "answer any question", since in dimensional modeling you have to decide ahead of time the nature of the questions that
will be asked. For the same reason, data mining is better supported by 3NF rather than dimensional models.
Most data warehouses projects which adopt dimensional modeling eventually find they have to support an additional ODS (operational data
store) of the detailed data that produced the dimensional data in order to answer the detailed questions that arise after general questions
are answered by the dimensional model.