Archives of the TeradataForum

Message Posted: Mon, 22 Aug 2005 @ 08:18:43 GMT

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

          > Attribute = Char(10), NOT NULL
          > Table > 100M Rows
          > Distinct Values = 6
          > Attribute is a secondary index on the table

I don't think I have performance stats you are looking for. Perhaps the best way to get them is to conducts a series of tests with a copy of the table on your own system?

By the way, I wonder whether such an index is worth keeping at all. if the distribution of values is even then each value is present in more than 15% of the rows. I think the optimizer is not likely to choose the SI because full table scan is more efficient in this case. If the distribution is not even and you have, say, one or two values present in only 5% of the rows, then the SI is probably still too "expensive" to keep.

If you decide to go for the series of tests on your own system, you might add a scenario with the SI dropped. I would not be surprised if it proves to be a winner.

Regards,

Victor

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