When I saw the code drop notice, that was my first thought
also. After some very nice help from the IBM lab in Boeblingen,
the answer is no.

They have implemented the CPU timer for a monitoring function,
but it has nothing to do with the accounting function.  It's
being used to decide when to update the monitor record. So the
numbers produced will still be bad as designed, and the monitor
records themselves will be suspect.

The implementation of the monitor record function to my thinking
is "low cost", but invalid. When I do performance analysis,
I really want to be able to correlate data. This implementation
has significant problems in this area, ok for capacity planning,
but not overly useful for performance analysis. And the implementation
is unique to VM - and the plans to populate this monitor record
will not help Michael in his needs.

>Date:         Tue, 2 Mar 2004 12:40:51 -0500
>From: "Ferguson, Neale" <[EMAIL PROTECTED]>
>
>Barton,
>
>Will the latest stuff in the January code drop help here?  The
>combination of the virtual timer support (which is based on the
>CPU timer) and the monitor data production by the kernel should
>allow more accurate statistics to be produced.







"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