Windows 2003 No page file - Pages/Sec

  • Thread starter Thread starter Justin Rich
  • Start date Start date
J

Justin Rich

So i have a rather beefy server that ive taken the page file off of. when i
run perfmon and look at pages/sec its HIGH... like 2000-8000

since i have no page file what exactly is this measuring?

Thanks
Justin
 
And what happens when you put a pagefile back on? In most cases running
without a pagefile is counterproductive.

John

Justin Rich wrote:

> So i have a rather beefy server that ive taken the page file off of. when i
> run perfmon and look at pages/sec its HIGH... like 2000-8000
>
> since i have no page file what exactly is this measuring?
>
> Thanks
> Justin
>
>
 
without it im seeing on average about 100.. with spikes here and there.

the system has 16GB of ram, with typically about 4GB in use.
also its a 32bit OS with PAE enabled.

Can you explain what you mean by counter productive?

Thanks
Justin

"John John" <audetweld@nbnet.nb.ca> wrote in message
news:Ojr2dX22HHA.4184@TK2MSFTNGP06.phx.gbl...
> And what happens when you put a pagefile back on? In most cases running
> without a pagefile is counterproductive.
>
> John
>
> Justin Rich wrote:
>
>> So i have a rather beefy server that ive taken the page file off of. when
>> i run perfmon and look at pages/sec its HIGH... like 2000-8000
>>
>> since i have no page file what exactly is this measuring?
>>
>> Thanks
>> Justin
 
It went from 8000 faults to 100, that pretty well explains
"counterproductive". Something making repeated hits on a non existent
pagefile is counterproductive. With the pagefile present these hits
almost disappear. When applications start they may ask for *way* more
RAM than they need, in fact most do ask for way more than needed.
Instead of simply giving the apps all that they immediately ask for the
memory manager will give the application what it (the Memory Manager)
thinks is adequate and map the remainder of the request to the paging
file. That doesn't mean that the paging file will be used, it just
makes the application think it got all it asked for and shuts up the
application from asking for more memory than it really needs. If while
working the application does actually need more RAM then it will not
necessarily page, if sufficient memory is available the Memory Manager
will simply take unused memory pages from another working set and give
it to the application making the request for additional RAM, when that
happens the mapping to the pagefile will once again change. The
information here may explain things a bit more:

PerfMon: High Number of Pages/Sec Not Necessarily Low Memory
http://support.microsoft.com/kb/139609

Think of it as an 8 year old kid asking for $100, way more than he needs:

Kid: "Can I have a hundred bucks, Dad?"
Dad: "No"
Kid: "Can I have a hundred bucks, Dad?"
Dad: "No"
Kid: "Can I have a hundred bucks, Dad?"
Dad: "No"

This will go on ad infinitum until Dad opens his wallet and gives the
kid $5. The kid is happy and he shuts up, he didn't really need $100,
$5 is enough. And he knows that there is a wallet full if he spends all
of the $5 he just got, he thinks he has $100 because he saw that Dad's
wallet was full. The kid is happy he got what he needs and Dad is happy
too, the kid stopped his incessant request and there is money left in
the wallet for other necessary things.

John

Justin Rich wrote:

> without it im seeing on average about 100.. with spikes here and there.
>
> the system has 16GB of ram, with typically about 4GB in use.
> also its a 32bit OS with PAE enabled.
>
> Can you explain what you mean by counter productive?
>
> Thanks
> Justin
>
> "John John" <audetweld@nbnet.nb.ca> wrote in message
> news:Ojr2dX22HHA.4184@TK2MSFTNGP06.phx.gbl...
>
>>And what happens when you put a pagefile back on? In most cases running
>>without a pagefile is counterproductive.
>>
>>John
>>
>>Justin Rich wrote:
>>
>>
>>>So i have a rather beefy server that ive taken the page file off of. when
>>>i run perfmon and look at pages/sec its HIGH... like 2000-8000
>>>
>>>since i have no page file what exactly is this measuring?
>>>
>>>Thanks
>>>Justin
>
>
>
 
I think the link you provided clears it up for me.

the application we run is basically just a huge compressor.. so i assume
when i see a page its actually just pulling in new data, rather than using a
page file... since its pretty much constantly new data i should be seeing a
lot of this.

Thanks for the info!

Justin

"John John" <audetweld@nbnot.nb.ca> wrote in message
news:O5PBPJ42HHA.5980@TK2MSFTNGP04.phx.gbl...
> It went from 8000 faults to 100, that pretty well explains
> "counterproductive". Something making repeated hits on a non existent
> pagefile is counterproductive. With the pagefile present these hits
> almost disappear. When applications start they may ask for *way* more RAM
> than they need, in fact most do ask for way more than needed. Instead of
> simply giving the apps all that they immediately ask for the memory
> manager will give the application what it (the Memory Manager) thinks is
> adequate and map the remainder of the request to the paging file. That
> doesn't mean that the paging file will be used, it just makes the
> application think it got all it asked for and shuts up the application
> from asking for more memory than it really needs. If while working the
> application does actually need more RAM then it will not necessarily page,
> if sufficient memory is available the Memory Manager will simply take
> unused memory pages from another working set and give it to the
> application making the request for additional RAM, when that happens the
> mapping to the pagefile will once again change. The information here may
> explain things a bit more:
>
> PerfMon: High Number of Pages/Sec Not Necessarily Low Memory
> http://support.microsoft.com/kb/139609
>
> Think of it as an 8 year old kid asking for $100, way more than he needs:
>
> Kid: "Can I have a hundred bucks, Dad?"
> Dad: "No"
> Kid: "Can I have a hundred bucks, Dad?"
> Dad: "No"
> Kid: "Can I have a hundred bucks, Dad?"
> Dad: "No"
>
> This will go on ad infinitum until Dad opens his wallet and gives the kid
> $5. The kid is happy and he shuts up, he didn't really need $100, $5 is
> enough. And he knows that there is a wallet full if he spends all of the
> $5 he just got, he thinks he has $100 because he saw that Dad's wallet was
> full. The kid is happy he got what he needs and Dad is happy too, the kid
> stopped his incessant request and there is money left in the wallet for
> other necessary things.
>
> John
>
> Justin Rich wrote:
>
>> without it im seeing on average about 100.. with spikes here and there.
>>
>> the system has 16GB of ram, with typically about 4GB in use.
>> also its a 32bit OS with PAE enabled.
>>
>> Can you explain what you mean by counter productive?
>>
>> Thanks
>> Justin
>>
>> "John John" <audetweld@nbnet.nb.ca> wrote in message
>> news:Ojr2dX22HHA.4184@TK2MSFTNGP06.phx.gbl...
>>
>>>And what happens when you put a pagefile back on? In most cases running
>>>without a pagefile is counterproductive.
>>>
>>>John
>>>
>>>Justin Rich wrote:
>>>
>>>
>>>>So i have a rather beefy server that ive taken the page file off of.
>>>>when i run perfmon and look at pages/sec its HIGH... like 2000-8000
>>>>
>>>>since i have no page file what exactly is this measuring?
>>>>
>>>>Thanks
>>>>Justin
>>
>>
>>
>
 
The Performance Counter called "Pages/Sec" counts a lot of memory related
events that are not related to the use or none use of a page file. For
example, Pages/Sec counter includes count of "Demand Zero Page Faults"
which is a count of requests from applications for new, empty, virtual
memory pages.

See
http://members.shaw.ca/bsanders/WindowsGeneralWeb/RAMVirtualMemoryPageFileEtc.htm,
particularly section 2.8 Monitoring RAM and Virtual Memory usage.

I'm not sure that eliminating the page file makes a lot of sense. If the
computer has enough RAM to satisfy the Virtual Memory demands of the
applications, the page file won't be used very much anyway.

--
Bruce Sanderson MVP Printing
http://members.shaw.ca/bsanders

It is perfectly useless to know the right answer to the wrong question.



"Justin Rich" <jrich523@yahoo.spam.com> wrote in message
news:uX$u$m12HHA.5164@TK2MSFTNGP05.phx.gbl...
> So i have a rather beefy server that ive taken the page file off of. when
> i run perfmon and look at pages/sec its HIGH... like 2000-8000
>
> since i have no page file what exactly is this measuring?
>
> Thanks
> Justin
>
 
Great website. Glad you made it :)

Well, i had orginally set the page file to 4GB which you would think would
be enough considering there is 16GB of RAM in the box but apparently it
still felt the need to grow... upwards of 22GB... seems a touch large to
me... and i have yet to see a process leak or go out of control in any way.

I have seen a lot of people say "you should have one" but i have yet to see
why. if there is enough memory then why do i need a page file? the system
only ever uses about 4GB (what a waste) and yet the page file still gets
nailed..

my reason as to why this is the case is that MS uses the page file way more
than it should as a security precaution.

If i have all of this memory why on earth would i want a page file? there is
more memory in there than im even able to use.

Thanks
Justin

"Bruce Sanderson" <bsanders@newsgroups.nospam> wrote in message
news:%2358RBFj3HHA.5796@TK2MSFTNGP05.phx.gbl...
> The Performance Counter called "Pages/Sec" counts a lot of memory related
> events that are not related to the use or none use of a page file. For
> example, Pages/Sec counter includes count of "Demand Zero Page Faults"
> which is a count of requests from applications for new, empty, virtual
> memory pages.
>
> See
> http://members.shaw.ca/bsanders/WindowsGeneralWeb/RAMVirtualMemoryPageFileEtc.htm,
> particularly section 2.8 Monitoring RAM and Virtual Memory usage.
>
> I'm not sure that eliminating the page file makes a lot of sense. If the
> computer has enough RAM to satisfy the Virtual Memory demands of the
> applications, the page file won't be used very much anyway.
>
> --
> Bruce Sanderson MVP Printing
> http://members.shaw.ca/bsanders
>
> It is perfectly useless to know the right answer to the wrong question.
>
>
>
> "Justin Rich" <jrich523@yahoo.spam.com> wrote in message
> news:uX$u$m12HHA.5164@TK2MSFTNGP05.phx.gbl...
>> So i have a rather beefy server that ive taken the page file off of. when
>> i run perfmon and look at pages/sec its HIGH... like 2000-8000
>>
>> since i have no page file what exactly is this measuring?
>>
>> Thanks
>> Justin
>>
>
 
Back
Top