Virtual memory problem? - PF usage grows, not freed when apps clos

[Guest]

Gerry,

Thanks for the discusion, information, and suggestions. It looks like the
problem is fixed.

The page file use now grows only as apps are added, and shrinks again as
apps are closed. It looks like the Tablet PC (TCServer) leak was the culprit,
and the patch seems to have taken care of the problem.

BTW, I have added the additional RAM (now at 1 GB) as planned. I just didn't
want to mask the problem, so I was waiting until I felt that was resolved.
Certainly it's best of Windows doesn't need to use the PF/disk

I have indeed heard the debate about 'optimal' VM settings... IMHO, in the
case of a system with small RAM and multiple apps in use simultaneously, I
think it made sense to allocate extra PF space (it appears that Window's
attempt to increase it when needed were less than ideal).

I'm curious about one aspect of this debate - I'm not clear on what would be
the down side of a larger maximum size? Does Windows use more PF space if it
seems to be available? Sounds strange, but otherwise what's the harm in
allocating a large inital or maximum setting (if the extra GB or two of disk
space is not an issue)?

Thanks much for the help!

cheers,
Gregg

 

[deebs]

Gerry,

Thanks for the discusion, information, and suggestions. It looks like the
problem is fixed.

The page file use now grows only as apps are added, and shrinks again as
apps are closed. It looks like the Tablet PC (TCServer) leak was the culprit,
and the patch seems to have taken care of the problem.

BTW, I have added the additional RAM (now at 1 GB) as planned. I just didn't
want to mask the problem, so I was waiting until I felt that was resolved.
Certainly it's best of Windows doesn't need to use the PF/disk

I have indeed heard the debate about 'optimal' VM settings... IMHO, in the
case of a system with small RAM and multiple apps in use simultaneously, I
think it made sense to allocate extra PF space (it appears that Window's
attempt to increase it when needed were less than ideal).

I'm curious about one aspect of this debate - I'm not clear on what would be
the down side of a larger maximum size? Does Windows use more PF space if it
seems to be available? Sounds strange, but otherwise what's the harm in
allocating a large inital or maximum setting (if the extra GB or two of disk
space is not an issue)?

Thanks much for the help!

cheers,
Gregg

This is of interest to me as well.

I think there are upsides to declaring Max = Min = Windows recommended
value on multiple platter arrangements.

It seems that Windows manages these so, in my case, I assigned 1533 MB
to two platters Max = Min option chosen.

On C: I set it as Windows system managed = 1533 MB with the hope that
this declaration forces Windows to manage the other 2 pagefiles as well
(it seems to)

Advantages?
- setting a fixed Max = Min seems to avoid defrag issues

- Windows seems to spread the page file usage across all three declared
areas (thereby optimising read/write stuff to drives?)

- all programs seem zippier and zappier

I'd speculate that a series based approach is probably OK but there
seems to be some parallel functioning on the go that is swifter than a
more powerful machine using a series based approach (vagueness intended)

 

[Gerry Cornell]

Gregg

Alex Nichol's Article is as good on virtual memory settings as any:
http://aumha.org/win5/a/xpvm.htm

The main point is not to set a maximum too low if you feel the need to set
a
maximum. In reality you should add memory so that, except when the system
has abnormal demands for memory, it only uses minimal amounts of virtual
memory. In this situation the maximum is hardly going to be relevant. If
there
is no maximum set the reality becomes the amount of free disk space.

The problems start to arise, however, when the motherboard limits on
memory prevent the user adding memory to avoid over dependence on virtual
memory. Also in the situation that the hard disk utilisation gets to 85%
the user looks to cut back on space allocations and the pagefile size comes
into
the equation. The answer is to get more memory or a larger hard disk but the
ability of the user to afford to buy either can make for a difficult
balancing act.

I am not aware that there is a downside in terms of system performance with
a large maximum. However it is arguably wasteful in terms of utilisation of
available disk space. The argument for a reasonably sized minimum is to
attempt to minimise resizing of the pagefile. In practice I wonder how
significant the impact on performance is? Read / write speeds to memory
and to disk and CPU speed are much more significant.

--

Hope this helps.

Gerry
~~~~
FCA
Stourport, England

Enquire, plan and execute
~~~~~~~~~~~~~~~~~~~

 

[Guest]

...
I am not aware that there is a downside in terms of system performance with
a large maximum. ... In practice I wonder how
significant the impact on performance is? Read / write speeds to memory
and to disk and CPU speed are much more significant.

Presuming that you're referring to the impact of resizing one the maximum is
reached, on my system at least, the resulting behavior was NOT pretty.
Windows pretty much hung for some period of time.

By the way, I may have spoken to soon regarding the problem being entirely
fixed. It's much better, however after running for 3 days without rebooting,
pagefile usage was 650 MB AFTER closing all apps except TM.

Can you (or anyone) tell me whether this is normal for Windows (XP) - that
the PF usage isn't fully released after all apps are closed?

thanks,
Gregg

 

[Gerry Cornell]

Gregg

I do not know the answer to your last question.

Do you switch your computer off when not in use?

We found Tablet. Perhaps we should be looking for another?

--

Regards.

Gerry
~~~~
FCA
Stourport, England

Enquire, plan and execute
~~~~~~~~~~~~~~~~~~~

 

[Guest]

I do not know the answer to your last question.

What happens if you use Windows for a while, then close all tasks - is the
PF usage reduced below say, 300 MB?

Do you switch your computer off when not in use?

Nope. Clearly, switching off and rebooting periodically would resolve (or at
least work around) this behavior. I'd rather avoid this practice if possible.

I prefer to leave Windows up for several reasons:

1) I have the computer download virus updates, defrag, etc over night
2) Windows is far from "instant on"
3) I prefer to save my working "context" from day to day when possible

We found Tablet. Perhaps we should be looking for another?

I guess so (regretably ... it still seems to me that this behavior of
not releasing PF memory resembles an OS leak (or 'feature'?).

[ Since
1) the memory use starts normal after all startup processes are loaded, I
don't think it's excessive memory use by added background processes, and
2) the PF memory is not freed when all apps are closed. ]

Maybe I have a bad a ssumption though that PF use should go back to at least
somewhere _close_ to it's post-start up value after all apps closed. We
could be chaising a wild goose here. But I think it's worth looking a bit
further, since it seems to me that the PF use should drop below 600 MB with
everything closed.

By the way, in the course of this investigation, I have found that there are
at least 2 memory leaks in Firefox. (I do have the latest version - v1.5,
and I'm only using one, widely used extension - RoboForm). I have installed
the 'fix' for one of the leaks
(http://www.freerepublic.com/focus/f-bloggers/1327586/posts), and the other
appeard to have no fix as of yet - http://kb.mozillazine.org/Memory_Leak.

But this still doesn't explain (I don't think) why the PF use stays so much
higher than the initial, after all apps are closed.

Any further thoughts or suggestions?

 

[Gerry Cornell]

Gregg

Do you use Outlook Express? If yes do you have the option to
"Read all messages in Plain Text" selected -Tool, Options, Read.

These Knowledge Base Articles appertain to memory leaks in
Windows XP (we are assuming that is the problem ).

Computer Speed and Performance May Decrease
http://support.microsoft.com/default.aspx?scid=kb;en-us;310419

MFC applications leak GDI objects on computers that are running
Windows XP
http://support.microsoft.com/default.aspx?scid=kb;en-us;319740

Intermittent Program Unresponsiveness Occurs When You Use
Performance Monitoring
http://support.microsoft.com/default.aspx?scid=kb;en-us;330259

I have reservations that the solutions offered may deal with the symptoms
and not resolve the problem.

--

Hope this helps.

Gerry
~~~~
FCA
Stourport, England

Enquire, plan and execute
~~~~~~~~~~~~~~~~~~~

 

[Gerry Cornell]

Greg

Another small utility to monitor pagefile usage:
http://www.dougknox.com/xp/utils/xp_pagefilemon.htm

--

Hope this helps.

Gerry
~~~~
FCA
Stourport, England

Enquire, plan and execute
~~~~~~~~~~~~~~~~~~~

I do not know the answer to your last question.

What happens if you use Windows for a while, then close all tasks - is the
PF usage reduced below say, 300 MB?

Do you switch your computer off when not in use?

Nope. Clearly, switching off and rebooting periodically would resolve (or
at
least work around) this behavior. I'd rather avoid this practice if
possible.

I prefer to leave Windows up for several reasons:

1) I have the computer download virus updates, defrag, etc over night
2) Windows is far from "instant on"
3) I prefer to save my working "context" from day to day when possible

We found Tablet. Perhaps we should be looking for another?

I guess so (regretably ... it still seems to me that this behavior of
not releasing PF memory resembles an OS leak (or 'feature'?).

[ Since
1) the memory use starts normal after all startup processes are loaded, I
don't think it's excessive memory use by added background processes, and
2) the PF memory is not freed when all apps are closed. ]

Maybe I have a bad a ssumption though that PF use should go back to at
least
somewhere _close_ to it's post-start up value after all apps closed. We
could be chaising a wild goose here. But I think it's worth looking a bit
further, since it seems to me that the PF use should drop below 600 MB
with
everything closed.

By the way, in the course of this investigation, I have found that there
are
at least 2 memory leaks in Firefox. (I do have the latest version - v1.5,
and I'm only using one, widely used extension - RoboForm). I have
installed
the 'fix' for one of the leaks
(http://www.freerepublic.com/focus/f-bloggers/1327586/posts), and the
other
appeard to have no fix as of yet - http://kb.mozillazine.org/Memory_Leak.

But this still doesn't explain (I don't think) why the PF use stays so
much
higher than the initial, after all apps are closed.

Any further thoughts or suggestions?

 

[Guest]

Another small utility to monitor pagefile usage:
http://www.dougknox.com/xp/utils/xp_pagefilemon.htm

That's the page file monitor porgram I asked baout earlier. I have't checked
it out yet, but I will do so (after this weekend probably - I alerady have a
full weekend of (mostly) work)

Do you use Outlook Express? If yes do you have the option to
"Read all messages in Plain Text" selected -Tool, Options, Read.

I use the full version of Outlook. Maybe the same problem. I'd rather not
disable HTML e-mfunctionality, but can do so to check and see if that
(Outlook) is the (a? problem.

These Knowledge Base Articles appertain to memory leaks in
Windows XP (we are assuming that is the problem ).

Computer Speed and Performance May Decrease
http://support.microsoft.com/default.aspx?scid=kb;en-us;310419

Interesting - did you notice this bit?

"When you quit a program, the system resources that the program uses should
be returned to the operating system. However, some programs do not return all
of these resources, effectively "leaking" memory, and this can create a low
system-resource state."

Meaning (as I read it) that, if this happens Windows does not recover the
memory. So it could be an application problem, which Windows doesn't manage
or recover from well.

MFC applications leak GDI objects on computers that are running
Windows XP
http://support.microsoft.com/default.aspx?scid=kb;en-us;319740

Great... (and this is production quality code?!? Never mind ...)

Intermittent Program Unresponsiveness Occurs When You Use
Performance Monitoring
http://support.microsoft.com/default.aspx?scid=kb;en-us;330259

Likewise . Sigh ... (;-)

Thanks a lot for the pointers. I suppose I'll start by monitoring the
pagefile usage with the tool, and check for GDI object leakage. And probably
continue to reboot periodically . I'll let you know what I learn.

thanks again - cheers,
Gregg

 

[David Candy]

It will release what parts the program's memory will be put in to the swap. Swap files are faster than normal disk reads. Windows puts stuff in memory and won't release it, like fonts. This is because they are designed to be always loaded but so you don't have to wait an hour for system bootup it demand loads on first use. If not needed it gets swapped out where it can quickly be swapped in.

But you aren't using 650MB of swap. It is merely reserved. Type systeminfo to see what is being used.

 

[David Candy]

What do you mean the full version of Outlook. Outlook and Outlook Express have completely different lineages. Except OE provides OL with internet mail (as OL history is not in internet). Outlook Express is the full version and the outlook team had nothing to do with it. OE file name and previous name is Microsoft Internet Mail And News. OL is derivitive from Schedule +.

 

[Gerry Cornell]

Gregg

The problem arises if "Read all messages is selected" is selected not the
reverse!
Apparently it is a memory leak bug which currently has no fix, except you
can either
exit from Outlook Express or not select the Plain Text option. It has been
fixed in the
but the fix has been broken by subsequent Windows XP updates. The KB 319740
fix does not fix the Outlook Express leak.

There is another ( perhaps free fix ) from a third party source doing the
same as the KB 319740 fix.
http://www.cakewalk.com/support/kb/kb2005243.asp

BTW I would not get hung up on this problem if an occasional restart of a
faulting
programme circumvents the consequences of it becoming a real problem.


--

Hope this helps.

Gerry
~~~~
FCA
Stourport, England

Enquire, plan and execute
~~~~~~~~~~~~~~~~~~~