Re: SQL Tuning Regarding System CPU Stats

andrew.markiewicz_at_gmail.com wrote:
>
> Thanks Connor. That was a good sanity check. I've had a hard time
> finding any other info supporting my thoughts about a test system so I
> was beginning to wonder if I had irrational concerns.
>
> >Nothing make a IT shop look
> >more stupid than having something signed off in dev/test as "working
> >fine" only to have it bog down to a crawl in production.
>
> Precisely what I want to avoid.
>
> >we strive to have our dev/test environments running fill size
> >production copies wherever its possible to do so.
>
> Our db is small enough that we are able to make full clones in test,
> so some of the difficulty involving data volume and statistics
> disparity is removed.
>
> Do you find it necessary to have different system stats in production
> for OLTP and batch processing? If so, how do you setup test
> environments to work with both scenarios? Two db?
>
> Thanks
> Andrew

I think that's far too granular...I reckon you should change system stats very rarely.

We gather our system stats each weekday between 2 and 3pm (which is the busiest workload of the day), but I stress, this is only *gather* not *activate*. Every couple of months we plot the data and see if its worth actually activating a new set of stats.

But stability/predictability is king - of course the gathered stats change slightly from day to day, week to week, etc - but we last did an activation around 10 months ago when we upgraded some server components, and I don't see us planning another for sometime.

Cheers
Connor

-- 
Connor McDonald
Co-author: "Mastering Oracle PL/SQL - Practical Solutions"
Co-author: "Oracle Insight - Tales of the OakTable"

web: http://www.oracledba.co.uk
web: http://www.oaktable.net
email: connor_mcdonald_at_yahoo.com


"Semper in excremento, sole profundum qui variat."

------------------------------------------------------------