Re: System Statistics and Calibrate IO

From: Subodh Deshpande <deshpande.subodh_at_gmail.com>
Date: Mon, 5 Dec 2011 12:43:04 +0530
Message-ID: <CAJsOtB5q42Ng=N8NbnU15YYJC-zwOOxp_SsgvzNU5i2XBH0cGg_at_mail.gmail.com>



yes, I think its upgrade project then all the parameters such memory, cpu, hardware are not going to change
then what may change is only I/O..
look at it from customer requirement..so if you at least have some minimum stat about previous I/O you can justify the performance after upgrade..and post actions taken if required..

if you keep all the stat it may help you for your own learning experience..and if you are habitual with upgrades then perhaps may not be required to gather the stats..just depending upon the performance you can take the actions if any are required...

thanks..subodh

On 3 December 2011 06:10, Greg Rahn <greg_at_structureddata.org> wrote:

> I don't see a technical requirement to run either. I have yet to use
> system stats on any warehouse workload, and calibrate is only necessary for
> 11.2.0.2 auto DOP, and then I would just set the values manually - the same
> way the Exadata best practice guide recommends.
> On Fri, Dec 2, 2011 at 4:15 PM, Kellyn Pot'vin <kellyn.potvin_at_ymail.com
> >wrote:
>
> > I'm looking for some *extra* info before I pull my hair out on this one,
> > folks!
> > I'm managing and performing an 11g upgrade project and today, there was
> > concern raised about me adding the dbms_resource_manager.calibrate_io run
> > instead of gather_system_stats in a database build process. Over the last
> > couple years reading here and there, I had come to believe that
> > calibrate_IO was something of a replacement for gather_system_stats, but
> > after research today, I'm not so sure about that!
> >
> > I have the basics down and am quite comfortable with the technical specs
> > of the new calibrate IO feature, but still feel there may very well be
> some
> > great value to gathering system stats, (This is for a mart build, is run
> > once with a sufficient work load produced on the system at the initial
> > build time, never again for the life of the datamart...)
> >
> > Can someone tell me why I would want to run one over the other, run both
> > or where I'm missing some data here?
> >
>
>
> --
> Regards,
> Greg Rahn | blog <http://bit.ly/u9N0i8> | twitter <
> http://bit.ly/v733dJ> |
> linkedin <http://linkd.in/gregrahn>
>
>
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

-- 
=============================================
TRUTH WINS AT LAST, DO NOT FORGET TO SMILE TODAY
=============================================


--
http://www.freelists.org/webpage/oracle-l
Received on Mon Dec 05 2011 - 01:13:04 CST

Original text of this message