Archives of the TeradataForum

Message Posted: Mon, 13 Aug 2001 @ 09:45:52 GMT

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

Margus,

I am not sure I can describe this problem simply but it took us 3 months to get to the bottom of it.

When statistics are collected 2 values are created for each of the samples a mode and a maximum value.

The enhancement for LIKE operations was only partially implemented in that if the LIKE qualification was satisfied by either one of the statistics values the index would be used if not then it would not. For our table of some 20m rows there were about 400 samples in the statistics, we where using names. Having printed the statistics values the tests proved that a partial value matching a statistics value would generate a query using the NUSI (very good performance improvement) but no other value would.

I am sure you will see that this hit and miss method makes the enhancement useless for improving performance.

As this was a Partners enhancement we have raised it with them but I do not think there is a commitment to fully implement the functionality.

Regards,

Peter

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