Archives of the TeradataForum
Message Posted: Fri, 21 Feb 2003 @ 21:59:40 GMT
Subj: | | Re: Can it be any faster |
|
From: | | Fred W Pluebell |
Doug Drake wrote:
| 3.) Dup row check - related to 1 above, if you have many duplicate values of the PI and the table is Multiset a dup row check will be
performed. The higher degree of duplication will cause dup row checks to rise exponentially. You can disable this check by using a SET
table but be aware that duplicates will be accepted | |
A few corrections on this point:
a) SET tables require dup row check, MULTISET tables allow duplicates
b) FastLoad never loads duplicate rows (even if the table is MULTISET)
c) But FastLoad uses a local pre-sort to eliminate duplicate rows, so even with skewed PI / RowHash distribution, impact of
completely duplicate rows is seldom a major factor.
d) On the other hand, logging of rows with duplicate UPI values (but not 100% duplicate content) to the error table can have a
major impact.
|