 |
 |
Archives of the TeradataForum
Message Posted: Fri, 09 Mar 2012 @ 13:38:03 GMT
Subj: | | Re: Partitioning on different column to PI |
|
From: | | Frydryszak, Marek |
| If the customer number is truly unique and used for access and we want to use partitioning on a date / timestamp column, then the best
solution would be to combine both columns in the primary index and partition on the date column - is this correct?" | |
NO !
If customer number is used for access, after adding date to PI all 1-amp queries will be all-amp and you will face dramatic performance
degradation.
On the other hand changing UPI to non-unique PI will not be noticeable from performance point of view. The only disadvantage is loss of the
unique constraint.
| |