Archives of the TeradataForum

Message Posted: Fri, 11 Apr 2003 @ 17:22:02 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

If there is more than one row per block the optimizer is going to do a table scan. If you think about it, to use a NUSI you have to read the nusi subtable, then read the data block. If you use the nusi in the worst case you could have 2 io's per data block instead of 1 with a table scan.

The optimizer is smart enough to look at the stats for the data value you are using in the search. If the data is skewed the optimizer won't use the nusi for the frequently occuring values but may use it for the infrequently occuring values.

	<Prev	Next>		<<First	<Prev	Next>	Last>>