Archives of the TeradataForum
Message Posted: Sat, 14 Jan 2012 @ 22:41:21 GMT
<-- Anonymously Posted: Saturday, January 14, 2012 17:21 -->
I wondered what solutions might? be extant for 'real' data that has skewed.
Consider this data distribution
Value count (*) 2323232 12 million 3545 120K 34349 450K 3434 45K etc
If we had skewing caused by NULLS and 'default data' values we can perhaps try some random number solns to make it look better, on a join column.
What about 'real' data? ( the same column is joined and selected ) that causes 'lumpy' distribution causing rowhashes to run amuck when that data get hashed , causing skewed CPU, increasing Impact values and degrading performance .
|Copyright 2016 - All Rights Reserved|
|Last Modified: 27 Dec 2016|