Re: Determing requested parallel slaves
Date: Tue, 30 Jun 2009 11:43:55 -0700 (PDT)
Message-ID: <5335df2f-186e-42c8-9c1c-e0f190bbf51a_at_b9g2000yqm.googlegroups.com>
On 30 June, 14:13, Mark D Powell <Mark.Pow..._at_eds.com> wrote:
> It sounds like all you really want is to count the number of
> associated pqo sessions in use for a session at a given point in
> time. What about just joining v$px_session back to v$session to v$sql
> and sampleing this every minute or two throughout the day.
>
> Personally I think using hints to limit the PQO sessions applied on a
> statement level is a fairly good idea. Overuse of PQO can kill
> performance.
>
> HTH -- Mark D Powell --- Hide quoted text -
>
> -
Thanks Mark.
Agreed PQO overdone can be very detrimental. Looking at those systems now!
The SQL join back is an idea. Still got to get from that to tying in the SQL that is part of a particular delivery. The test databases tend to support multiple test runs, so most of what is running in a day is nothing to do with what is being delivered that week.
I have an SR open for this, but the early response from the analyst is that there are no dictionary views to be queried after the event so something less elegant is going to be the only answer.
Rob Received on Tue Jun 30 2009 - 13:43:55 CDT