Archives of the TeradataForum

Message Posted: Fri, 18 May 2012 @ 16:25:56 GMT


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

Subj:		Re: Performance between development and production systems

From:		Gorner, Tomas

Hi,

From my understanding, the problem is Parsing time, not AMP CPU time. Christer said that the EXPLAIN takes longer, not the execution of the query itself.

Look into DBC.QRYLOG, the TotalCPUTime column was split into AMPCPUTime and ParserCPUTime (hopefully, it was already done in v12. It is definately the case in v13).

If my assumption is correct, then the delay should be visible in ParserCPUTime column. That is basically the time spent on building the execution plan (EXPLAIN). In that case, you may try to play with some buffer settings in DBS Control.

Any lock related delays should be visible in AMPCPUTime, not in ParserCPUTime, which is the time a parser needs to build the execution plan.

Locks are placed when the execution plan is beign executed, not when it is beign built. So, locks should not be an issue when running EXPLAIN. Neither should the data volume play any role.

Just my 2 cents...

Regards,

Tomas Gorner