Archives of the TeradataForum

Message Posted: Fri, 20 Jul 2001 @ 18:26:32 GMT

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

Helen,

Rajiv is correct. Even if you only specify the table name in the collect stats command (i.e. no column or index info) the actual processing - i.e. table/index scan - is repeated for every set of statistics. So, if it takes 20 minutes to scan a table and you have stats on 5 columns then a 'collect stats' command at the table level will take 100 minutes. This is true for v2r3 and v2r4.

The only speed up that I know of is if you're on v2r4 AND you're collecting stats on multiple columns.

In this situation, run multiple concurrent 'collect stats' jobs, naming the specific columns/indexes etc. With v2r4, these jobs will run in parallel and will not make each other wait (under v2r3 they would). The only possible contention is right at the end when they update the dictionary, but as that doesn't take very long, it doesn't matter.

The other thing to check of course is what else was on the box when your job took 15hours? Was the collect stats jon blocked by something else accessing the same table? (i.e. normal sort of questions and possibilities).

Cheers,

Dave

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