[Orca-users] RE: Orcallator Data collection

[Robinson, Paul]

We really like the way data is presented with orca. Not just the graphs but
the nice little zipped up package that has all the days stats. The bean
counters like to be able to get the data in the columnar format provided by
orca and they run it through their own statistical package that they have
developed. I never even see the results of that. I just provide them with a
data feed. We do use sar for specific things but are interested in the
possibility of the orcallator leveraging the SE toolkit to provide the
numbers, zip them up into a neat package & even graph them. 

I am not sure of their use for specific processor info. 

Great! I am glad to hear that the SE toolkit provides hooks into these
parameters. That will greatly increase the odds of successfully integrating
them into the orcallator :) 

Paul



-----Original Message-----
From: john65 at pobox.com [mailto:john65 at pobox.com]
Sent: Wednesday, December 12, 2001 2:31 PM
To: 'orca-users at yahoogroups.com'
Subject: Re: [orca-users] Orcallator Data collection




On Wed, 12 Dec 2001, Robinson, Paul wrote:

>  Blair,
>
> I am forwarding the following questions from our technical infrastructure
> bean counters.
>
> > There are a couple of types of data that I don't believe Orcallator
gives
> > us.  I'm not sure how hard these would be to add to the tool.
> >
> > The first is CPU utilization by processor.  This can be useful in the
case
> > of tasks which cannot take advantage of multiple processors.

I think this only makes sense if you actually bind processes to specific
 processors (or processor groups) using psrset.



> >
> > The second type of data is process data, i.e., data which shows which
> > tasks/workloads are actually using the resources (CPU, memory).
> >

Isn't process accounting the more appropriate place to get this kind of
information?

The setoolkit also has some tools to extract this kind of data.



> Paul Robinson
> Austin Automation Center
> >
> >
>