Task Manager PF Usage makes no sense

[Martin X.]

FYI

I decided to really dig into Task Manager today and found a few things that
don't make a lot of sense.

I did a test by starting a Virtual PC, which increased the PF Usage by
pretty much the same amount as the decrease in Physical Memory > Available
(~539 MB). I had set up the VPC for 510 MB of RAM, so that's pretty close to
the 533 number, which makes sense.

So if physical memory is decreased by an appropriate amount for a new
process, why is PF Usage also increased by roughly that same amount? This is
totally contradictory behavior and makes no sense to me. Based on my
findings, I say that PF Usage does not actually indicate page file usage, at
least not accurately by any means.

Another odd thing is that under the Process tab, VPC was only listed as
using 18 MB of memory. And nothing else listed under the processes accounted
for the ~533 MB used by the VPC. Yes, I did check the box to show processes
for all users.

 

[John John]

The PF figures shown in Task Manager are deceptive, use Perfmon to
accurately monitor the pagefile.

John

 

[Guest]

I'll second what John-John wrote and add the following:

The increase you saw in Taskmon was an allocation that was based on the
resources you spec'ed.

If you make a group reservation at a restaurant and you don't have the final
headcount, but believe it could go as high as 20 people, the restaurant is
going to set aside seating for twenty so that it's available for as many
people as you believe MAY need it. Perhaps only eleven people will show up,
perhaps twenty will. There's room for whatever portion of twenty bodies use
it.

The best way to REALLY see what's going on is to watch the behavior of one
machine from another. Each time you go in to TASKMAN, the activity level
spikes. The same is true each time you launch another app, another device or
service is called, etc. A lot of that "activity" is due to TASKMAN's own
overhead and the fact that it's the "focus" app which means it's getting a
few more cycles prioritized to it than a remote monitoring of the same
machine would.

As John-John said, TASKMAN's thermometers aren't the most reliable readings
to go by. A well-constructed audit using PERFMON, either locally or remotely,
will give you exponentially more reliable metrics.

Hope this helps.

Sam French