My default for 'correcting' the linux processor numbers is to
use TOTAL CPU time used by the virtual machine, but there is an
option if the installation wishes to use Virtual time.

Accuracy is based on one minute granularity. Very accurate to
show the total linux utilization, slightly less accurate when
analyzing individual processes, maybe off by as much as a couple
percent under some situations???  But much closer than the
potential order of magnitude as reported by TOP and other linux
tools....  When i used data granularities of an hour, i could
see larger discrepancies. One minute is easy and cheap....

>From:         Rob van der Heij <[EMAIL PROTECTED]>
>At 14:59 22-07-02 -0700, Barton Robinson wrote:
>
>>ESALPS will also provide a "multiplier" so that you can
>>correct your accounting data.
>
>That's a neat trick you can do when you have both numbers
>available, as ESALPS does. Is there any way you can guess
>how good the corrected numbers are?
>I suppose you correct based on consumed cycles in emulation
>mode because Linux will see that as its 100%, but somehow
>you also need to spread the cycles in supervisor mode over
>that to show the real cost?
>
>Rob







"If you can't measure it, I'm Just NOT interested!"(tm)

/************************************************************/
Barton Robinson - CBW     Internet: [EMAIL PROTECTED]
Velocity Software, Inc    Mailing Address:
 196-D Castro Street       P.O. Box 390640
 Mountain View, CA 94041   Mountain View, CA 94039-0640

VM Performance Hotline:   650-964-8867
Fax: 650-964-9012         Web Page:  WWW.VELOCITY-SOFTWARE.COM
/************************************************************/

Reply via email to