.Help with measuring working-set

.Help with measuring working-set

[Suresh Duddi]

hi, I am developer of Mozilla (open source web browser from mozilla.org) 
We are trying to make footprint improvements to the browser and have 
settled on minimizing working set and max-vm-usage as our goals.

http://www.mozilla.org/projects/footprint/footprint-guide.html

One thing we are struggling with is measurement of working set of app 
during a time interval.

Any pointers ? Are the metrics the best ones to measure and optimize ?

This is my first post to this mailing list; deeply sorry if this 
question is outside the agenda of this group.

dp

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: .Help with measuring working-set

[Andi Kleen]

On Mon, Feb 04, 2002 at 06:21:00PM -0800, Suresh Duddi wrote:
> hi, I am developer of Mozilla (open source web browser from mozilla.org) 
> We are trying to make footprint improvements to the browser and have 
> settled on minimizing working set and max-vm-usage as our goals.
> 
> http://www.mozilla.org/projects/footprint/footprint-guide.html
> 
> One thing we are struggling with is measurement of working set of app 
> during a time interval.

The only metric the kernel supports for the working set currently is a 
single RSS integer telling you have many pages are currently mapped into
a process. There may be more pages from your process unmapped in RAM, 
but being unmapped is usually the first step to swap the page out or 
throw it away. 

> Any pointers ? Are the metrics the best ones to measure and optimize ?

I guess you would prefer to know which pages are mapped at a given
point. This would require some custom patching to add a trace facility
for that. Shouldn't be that hard to implement, but I don't know of a 
ready patch.

-Andi
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: .Help with measuring working-set

[bert hubert]

On Wed, Feb 06, 2002 at 09:04:30AM +0000, Andi Kleen wrote:

> > Any pointers ? Are the metrics the best ones to measure and optimize ?
> 
> I guess you would prefer to know which pages are mapped at a given
> point. This would require some custom patching to add a trace facility
> for that. Shouldn't be that hard to implement, but I don't know of a 
> ready patch.

mincore(2) perhaps?

-- 
http://www.PowerDNS.com          Versatile DNS Software & Services
http://www.tk                              the dot in .tk
Netherlabs BV / Rent-a-Nerd.nl           - Nerd Available -
Linux Advanced Routing & Traffic Control: http://ds9a.nl/lartc
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: .Help with measuring working-set

[Rik van Riel]

On Wed, 6 Feb 2002, Andi Kleen wrote:
> On Mon, Feb 04, 2002 at 06:21:00PM -0800, Suresh Duddi wrote:

> > http://www.mozilla.org/projects/footprint/footprint-guide.html
> >
> > One thing we are struggling with is measurement of working set of app
> > during a time interval.

> > Any pointers ? Are the metrics the best ones to measure and optimize ?
>
> I guess you would prefer to know which pages are mapped at a given
> point. This would require some custom patching to add a trace facility
> for that. Shouldn't be that hard to implement, but I don't know of a
> ready patch.

I think Mike Shaver (you know him) has made a kernel patch
to measure exacty this, also for Mozilla development.

You should be able to just use his patch, if he still has
it.

regards,

Rik
-- 
"Linux holds advantages over the single-vendor commercial OS"
    -- Microsoft's "Competing with Linux" document

http://www.surriel.com/		http://distro.conectiva.com/

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/