Archives of the TeradataForum

Message Posted: Wed, 02 Feb 2012 @ 01:09:29 GMT

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

<-- Anonymously Posted: Wednesday, February 01, 2012 18:51 -->

Hi Dave and Machha,

Thanks for the questions.

That query ran once a day on different sessions, so the reason should not be caching. The data are identical and result set and explain are the same since we use it as a benchmark query to test system performance. Yes the AMPCPU etc are read from DBQLogTbl and DBQLSQLtbl.

We used to use it to compare the performance pre and post upgrade. However I just realised with the same configuration, the AMPCPU consumption has changed a lot, i.e. over 20%. I understand the elapsed time can be various based on system busy or not. But don't know why CPU can change that much on a same query at same patch level. The ParserCPU are the same.

We upgrade to TD13.10 recently on 5550 platform. It runs on ONE DIMENSIONAL MODE instead of TRADITIONAL MODE since upgrade so AMPs can share disk storage across arrays. The reason for that was in our previous hardware upgrade for some reason the disk distribution were not evenly built across AMPs.

Thanks

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