Archives of the TeradataForum

Message Posted: Fri, 12 Oct 2001 @ 13:32:39 GMT

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

Is COLLECT STATISTICS really always desirable?

I have a different experience.

A query which took to complete about two minutes, took more than an hour after stats have been collected on one of the tables involved.

Statitsics were collected on a collumn named DWE_Date with a rather uneven distribution (details see below). In the query the table is joined to a one-row table via this column.

The parser underestimates the size of the result of the join and duplicates on AMPs what it shoud not duplicate.

Petr Horsky

Details: 51157 rows, 207 different values, six of them with about 8 460 occurences, the others with 1-3 occurences

        DWE_Date        Count(*)
        2001-10-10      8 468
        2001-10-09      8 466
        2001-10-08      8 462
        2001-10-05      8 460
        2001-10-04      8 458
        2001-10-03      8 460
        2001-10-01      1
        2001-09-27      2
        2001-09-26      2
        2001-09-25      1
        etc.

joined to a one-row table MD_DWE_Date which holds the current date ie. 2001-10-10 (and then to some other tables).

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