cluster size of pagefile partition

[Timothy Daniels]

I plan to use a small partition on a 2nd hard drive
for the pagefile.

Is there a recommended cluster size for it?

How does one cause the pagefile's partition
to be on the outermost cylinders of the hard
drive?


Thanks for any info.
*TimDaniels*

 

[Rod Speed]

I plan to use a small partition on a 2nd hard drive for the pagefile.

Is there a recommended cluster size for it?

This is getting to obsessively anal again.

The thing to do first is to ensure that the system has
enough physical ram to minimise the use of the pagefile.

How does one cause the pagefile's partition to
be on the outermost cylinders of the hard drive?

Basically with a defragger that allows you to control that,
particular with fixed page files, as opposed to page files
that are allowed to shrink and grow as the need arises.

But again, it makes much more sense to have enough
physical ram so the pagefile isnt used much.

 

[Timothy Daniels]

"Rod Speed" opined:>

This is getting to obsessively anal again.

The thing to do first is to ensure that the system has
enough physical ram to minimise the use of the pagefile.

My physical RAM is maxed out. Is there a "best" or
recommended cluster size for a pagefile in a dedicated
partition?

Basically with a defragger that allows you to control that,
particular with fixed page files, as opposed to page files
that are allowed to shrink and grow as the need arises.

A de-fragger can control a *partition's* location?
Which de-fragger can do that?

But again, it makes much more sense to have enough
physical ram so the pagefile isnt used much.

But again, my RAM is maxed out.


*TimDaniels*

 

[J.Clarke]

"Rod Speed" wasted bandwidth:>


I have 2 hard drives. The 2nd hard drive lives
inside the PC. It is there. It is not going away. It
has lots of room. There is lots of room on it for a
partition to devote to a pagefile. I do not want to
buy or make another computer. Is there an optimal
or advisable format for the pagefile partition given
that the rest of the system is formatted NTFS?

No matter what you do to it a disk is thousands of times slower than
RAM--fiddling the page file is going to have a tiny effect on
performance when compared to putting in sufficient RAM.

Put the page file on the _first_ partition on the drive and don't use
that drive (not that partition, that whole drive) for _anything_ else
and you've pretty much done what you can.

I said at the start that the partition is dedicated to the
pagefile.





The 2nd hard drive, where the pagefile will be, will see
virtually no other use in normal operation.




The 2nd hard drive will be on its own channel on
an Ultra ATA 133 controller card.




More RAM, as I've repeatedly explained, is not an
option.

We are all curious--why not?

But again, the only thing that makes any sense
is to replace the motherboard so it isnt anymore.
And get a big bonus with an increased FSB
speed, and get a decent power supply instead
of the puny weakling you currently have as well.

[sic]



*TimDaniels*

 

[Rod Speed]

Rod Speed wrote

My physical RAM is maxed out.

Then the only thing that makes any sense is to
replace the PC so you can have more physical
ram and a much better power supply etc as well.

No obsessive fiddling with the page file detail will produce
anything like as good a return performance wise.

And you'll get a noticeable improvement
from the increased FSB speed as well.

Is there a "best" or recommended cluster
size for a pagefile in a dedicated partition?

There can never be any such animal.

It varys with the file structure used, FAT32
or NTFS etc, and that other factor of a fixed
pagefile or a dynamic one matters as well.

And its all mindlessly anal when it makes a
hell of a lot more sense to cut to the chase
and replace the system and have enough
physical ram so the page file isnt used at all.

A de-fragger can control a *partition's* location?

Nope. The location of the partition is set a partition creation time.

But unless you have a dedicated partition for the page file,
you would normally want to control where the page file is
in the partition if you're getting that anal about page files.

And even a dedicated partition for the page file has its
own downsides because if its on the same physical drive
as the partition which is seeing the most disk activity, the
heads can end up moving much further between the two
partitions when the system is excessively paging.

If you want to get obsessive about optimising page files,
you really need the page file on a different physical drive
to the one which is seeing the disk activity, and it needs
to be on a different controller too, and it cant be some
dinosaury old small drive, because its performance
may well be rather poor just because its a dinosaur too.

And when you are getting that obsessive about the page
file, it makes a hell of a lot more sense to do whatever
it takes to have enough physical ram so the page file
isnt used at all, because thats an order of magnitude
or more faster than any page file on a drive can ever be.

But again, my RAM is maxed out.

But again, the only thing that makes any sense
is to replace the motherboard so it isnt anymore.
And get a big bonus with an increased FSB
speed, and get a decent power supply instead
of the puny weakling you currently have as well.

Even a desperate pov wouldnt have to flip too
many burgers to be able to get a decent system.

 

[Timothy Daniels]

"Rod Speed" wasted bandwidth:>

Then the only thing that makes any sense is to
replace the PC so you can have more physical
ram and a much better power supply etc as well.

No obsessive fiddling with the page file detail will produce
anything like as good a return performance wise.

I have 2 hard drives. The 2nd hard drive lives
inside the PC. It is there. It is not going away. It
has lots of room. There is lots of room on it for a
partition to devote to a pagefile. I do not want to
buy or make another computer. Is there an optimal
or advisable format for the pagefile partition given
that the rest of the system is formatted NTFS?

But unless you have a dedicated partition for the page file,
you would normally want to control where the page file is
in the partition if you're getting that anal about page files.

I said at the start that the partition is dedicated to the
pagefile.

And even a dedicated partition for the page file has its
own downsides because if its on the same physical drive
as the partition which is seeing the most disk activity, the
heads can end up moving much further between the two
partitions when the system is excessively paging.

The 2nd hard drive, where the pagefile will be, will see
virtually no other use in normal operation.

If you want to get obsessive about optimising page files,
you really need the page file on a different physical drive
to the one which is seeing the disk activity, and it needs
to be on a different controller too, and it cant be some
dinosaury old small drive, because its performance
may well be rather poor just because its a dinosaur too.

The 2nd hard drive will be on its own channel on
an Ultra ATA 133 controller card.

And when you are getting that obsessive about the page
file, it makes a hell of a lot more sense to do whatever
it takes to have enough physical ram so the page file
isnt used at all, because thats an order of magnitude
or more faster than any page file on a drive can ever be.

More RAM, as I've repeatedly explained, is not an
option.

But again, the only thing that makes any sense
is to replace the motherboard so it isnt anymore.
And get a big bonus with an increased FSB
speed, and get a decent power supply instead
of the puny weakling you currently have as well.

[sic]



*TimDaniels*

 

[Rod Speed]

Rod Speed wrote

I have 2 hard drives.

Completely irrelevant. The only thing that makes any sense is to
replace the PC so you can have more physical ram if its using the
page file a lot now, and a much better power supply etc as well.

The 2nd hard drive lives inside the
PC. It is there. It is not going away.

It should. Into the bin.

It has lots of room.

What matters is whether that hard drive is a decent
modern high performance drive, if you're too stupid
to add more physical ram, however its necessary to
do that, if the page file is currently getting used much.

There is lots of room on it for a
partition to devote to a pagefile.

There is a hell of a lot more involved than just room, child.

I do not want to buy or make another computer.

More fool you.

Is there an optimal or advisable format
for the pagefile partition given that the
rest of the system is formatted NTFS?

Nope, it depends entirely on the detail of that hard
drive. Whether its on its own channel, whether its
some old dinosaur with lousy performance, etc etc etc.

And all of that is completely stupid compared with
adding more physical ram, however its necessary to
do that, if the page file is currently getting used much.

I said at the start that the partition is dedicated to the pagefile.

Thats not neccessarily the best approach either. If that
physical drive is being actively used at the same time as
the pagefile is being used, because you are stupid enough
to use a pagefile instead of more physical ram, you can
get better results with a fixed pagefile in the same partition
as the other drive activity. Just because the heads dont have
to move between two sets of partition directory structures etc.

And NTFS isnt necessarily the most appropriate format
for a dedicated partition for a page file anyway, because its
optimised for data security and you dont need that for a pagefile.

The 2nd hard drive, where the pagefile will be,
will see virtually no other use in normal operation.

Even more likely that its a dinosaury old drive with lousy
performance thats useless for a high activity pagefile.

You may even find that the use of that old dinosaur for the
pagefile gives significantly worse performance than having
a fixed pagefile on the main boot drive, particularly if the
main boot drive is a decent modern high performance drive.

The 2nd hard drive will be on its own channel
on an Ultra ATA 133 controller card.

It may well still have significant worse performance than
a fixed pagefile on the main boot drive, just because it
may well be some old dinosaur of a drive, and the total
performance of the drive subsystem may well be rather
worse than with a fixed pagefile on the main boot drive
if that boot drive is a decent modern high performance
drive with a decent bandwidth motherboard controller etc.

More RAM, as I've repeatedly explained, is not an option.

Its ALWAYS an option. And the only sensible one too.

[sic]

If you keep retching in public, you'll have to go to your room.

 

[Ron Martell]

My physical RAM is maxed out. Is there a "best" or
recommended cluster size for a pagefile in a dedicated
partition?

A 4K cluster size (which is the NTFS default for an aligned partition
and also the FAT32 default for partitions smaller than 8 gb) will be
an exact match for the 4K memory page size.

As some paging file activities will involve a single memory page this
is probably the optimum size. This is more relevant to Windows
95/98/Me where swap file operations are always done in 4K (single
page) blocks. Windows XP will aggregate page file reads and writes to
span multiple pages.

But as "Rod Speed" says the best situation for the page file is to
have enough RAM so as to eliminate or at least minimize the need to
actually write to it.

You can check on actual page file usage with a free utility from
http://www.dougknox.com/xp/utils/xp_pagefilemon.htm or from
http://billsway.com/notes_public/WinXP_Tweaks/

There is probably about 20 mb of stuff in RAM which can be moved to
the page file and which will rarely if ever be needed again. But if
the actual page file usage gets up to the 50 mb level or more on a
regular basis then that is pretty conclusive that there is
considerable paging in and out occurring and a RAM upgrade would
almost certainly be beneficial.

Note also that a dedicated page file partition can be
counter-productive in some instances. This is because much of the
paging file activity occurs interspersed with other disk activity such
as starting a new application program or opening a new data file. If
that is being done from a different partition on the same physical
drive as the paging file then the increased travel distance for the
drive head mechanism as it shuttles back and forth between the two
partitions can slow things down.

On the other hand, if the item being loaded is on a different physical
hard drive from the paging file then performance will be optimized
because the drive head mechanism can remain positioned at the page
file location.

Good luck



Ron Martell Duncan B.C. Canada
--
Microsoft MVP
On-Line Help Computer Service
http://onlinehelp.bc.ca

"The reason computer chips are so small is computers don't eat much."

 

[J.Clarke]

"J.Clarke" inquired:



Read the thread.

I see no messages in this thread in which you give any reason why more
RAM is not an option. In fact I can't find any post from you anywhere
that explains why.

My impression is that you're struggling to make an obsolete machine do
things that it is not capable of doing. Simple fact, if the machine
doesn't have enough RAM to do the job you need to do then you can tweak
the page file until Hell freezes over and it's still not going to have
enough RAM to do the job you need to do.

 

[Timothy Daniels]

"J.Clarke" inquired:

We are all curious--why not?

Read the thread.


*TimDaniels*

 

[Timothy Daniels]

A 4K cluster size (which is the NTFS default for an aligned partition
and also the FAT32 default for partitions smaller than 8 gb) will be
an exact match for the 4K memory page size.

As some paging file activities will involve a single memory page this
is probably the optimum size. This is more relevant to Windows
95/98/Me where swap file operations are always done in 4K (single
page) blocks. Windows XP will aggregate page file reads and writes to
span multiple pages.

But as "Rod Speed" says the best situation for the page file is to
have enough RAM so as to eliminate or at least minimize the need to
actually write to it.

You can check on actual page file usage with a free utility from
http://www.dougknox.com/xp/utils/xp_pagefilemon.htm or from
http://billsway.com/notes_public/WinXP_Tweaks/

There is probably about 20 mb of stuff in RAM which can be moved to
the page file and which will rarely if ever be needed again. But if
the actual page file usage gets up to the 50 mb level or more on a
regular basis then that is pretty conclusive that there is
considerable paging in and out occurring and a RAM upgrade would
almost certainly be beneficial.

On my current primary drive, the pagefile size is 576MB - 1.5
times the RAM size of 384MB. I plan to make the partition
on the 2nd hard drive the same size or larger. What is the
20MB "stuff" you mention "which can be moved to the pagefile"
and why is that "stuff" ever needed? I though a pagefile was just
virtual RAM, that is, just available work space.


*TimDaniels*

 

[Rod Speed]

On my current primary drive, the pagefile size
is 576MB - 1.5 times the RAM size of 384MB.

Mindlessly superficial way to size a pagefile.

In fact the more the physical ram, the smaller the
pagefile you can get away with most of the time.

I plan to make the partition on the 2nd hard drive
the same size or larger. What is the 20MB "stuff"
you mention "which can be moved to the pagefile"

Basically stuff thats not very likely to be used.

Used for unusual stuff.

and why is that "stuff" ever needed?

See above.

I though a pagefile was just virtual
RAM, that is, just available work space.

Its much more complicated than that with modern OSs.

Access to the pagefile is much quicker than to the individual
files that are part of the OS on the boot drive. Not just because
access to the pagefile is by blocks, no farting around with the
directory structures etc, but also because what that stuff is
loaded from can be expanded into the form actually used etc.

So it makes sense to have quite a bit of stuff in the page file
in case its ever needed, just because its quicker to get from
there. There is a real sense in which there is a tradeoff
between boot times and shorter app start times, particularly
when so much of what modern apps use comes out of dlls etc.

Its mostly the app start times that the
users notice when using a desktop PC.

 

[Timothy Daniels]

I see no messages in this thread in which you give
any reason why more RAM is not an option. In
fact I can't find any post from you anywhere
that explains why.

From my post in this thread on 9/10/03:
"My physical RAM is maxed out."

My impression is that you're struggling to make
an obsolete machine do things that it is not capable
of doing. Simple fact, if the machine doesn't have
enough RAM to do the job....

From my later post in this thread on the same day:

" I have 2 hard drives. The 2nd hard drive lives
inside the PC. It is there. It is not going away. It
has lots of room. There is lots of room on it for a
partition to devote to a pagefile. I do not want to
buy or make another computer. Is there an optimal
or advisable format for the pagefile partition given
that the rest of the system is formatted NTFS?"

Translation:

" I have 2 hard drives. The 2nd hard drive lives
inside the PC. It is there. It is not going away. It
has lots of room. There is lots of room on it for a
partition to devote to a pagefile. I do not want to
buy or make another computer. Is there an optimal
or advisable format for the pagefile partition given
that the rest of the system is formatted NTFS?"

In other words:

The physical facility is there, available at no cost.
I would like to use a small part of it for virual memory.

How does one best implement virtual memory on the
2nd hard drive in regards to format and cluster size?


*TimDaniels*

 

[Guess Who]

I've created a 1GB partition on my 2nd Seagate SATA 120GB drive. I created
this partition first, formatted NTFS with a cluster size of 32K, and use a
fixed size 850MB swap file. But I've got 2GB of RAM so I probably don't
need that big a swap file anyway.

I don't know if it's optimal, but it works for me.

 

[Timothy Daniels]

"Guess Who" announced:

I've created a 1GB partition on my 2nd Seagate
SATA 120GB drive. I created this partition first,
formatted NTFS with a cluster size of 32K, and
use a fixed size 850MB swap file. But I've got
2GB of RAM so I probably don't need that big
a swap file anyway.

I don't know if it's optimal, but it works for me.

Sounds good. Is your system partition NTFS
as well?


*TimDaniels*

 

[Ron Martell]

On my current primary drive, the pagefile size is 576MB - 1.5
times the RAM size of 384MB. I plan to make the partition
on the 2nd hard drive the same size or larger. What is the
20MB "stuff" you mention "which can be moved to the pagefile"
and why is that "stuff" ever needed? I though a pagefile was just
virtual RAM, that is, just available work space.

It is space on the hard drive that can, and is, used as a substitute
for RAM.

When the total RAM requirements exceed the amount of physical RAM then
something has to go. The Memory Manager tracks all items in RAM.
Some, generally critical core components of Windows, are flagged as
non-pageable meaning they must always be kept in RAM. But a good
proportion of the RAM content is eligible to be moved to the pagefile.

The memory manager keeps track of each 4K page of eligible memory,
noting how often it is used and when it was last used. When RAM is
needed for new, more important tasks, the memory manager chooses the
least used/longest inactive items already in memory and moves enough
of these out to the page file so that the RAM they were occupying can
be freed up for the new requirements.

Should it happen that something that was moved to the page file is
needed again by Windows or an application program then the memory
manager must identify something else that is inactive in RAM and move
it out to the page file and then move the require item back from the
page file into RAM so it can be used. That is the "swapping" process
that gives rise to the name swap file which is sometimes used as a
synonym for paging file.

My comment was meant to indicate that there seems to be about 20 mb
worth of code that is loaded by Windows at startup and which is quite
unlikely to be needed again, except perhaps at shutdown. The memory
manager can therefore move that code to the paging file when it needs
space and it will probably not have to be moved back in again. So the
first 20 mb or so of actual page file usage does not normally imply
that there is a lot of paging back and forth between RAM and the
paging file. These items are moved out once and seem to be left
there.

I have no idea what specific parts of the Windows code this
represents. I suspect it is items associated with initialisation and
possibly with error recovery.


Ron Martell Duncan B.C. Canada
--
Microsoft MVP
On-Line Help Computer Service
http://onlinehelp.bc.ca

"The reason computer chips are so small is computers don't eat much."

 

[Timothy Daniels]

My comment was meant to indicate that there seems
to be about 20 mb worth of code that is loaded by
Windows at startup and which is quite unlikely to be
needed again, except perhaps at shutdown. The memory
manager can therefore move that code to the paging file
when it needs space and it will probably not have to be
moved back in again. So the first 20 mb or so of actual
page file usage does not normally imply that there is a lot
of paging back and forth between RAM and the paging file.
These items are moved out once and seem to be left there.

I have no idea what specific parts of the Windows code
this represents. I suspect it is items associated with
initialisation and possibly with error recovery.

Thanks, Ron. Can this 20MB or so of pagefile requirement
be satisfied by the pagefile partition on the 2nd hard drive,
or must it be accommodated by pagefile space on the system
drive?


*TimDaniels*

 

[Ron Martell]

Thanks, Ron. Can this 20MB or so of pagefile requirement
be satisfied by the pagefile partition on the 2nd hard drive,
or must it be accommodated by pagefile space on the system
drive?

If there are multiple pagefiles configured then Windows will pick and
choose between them, deciding which is the most efficient for each
paging operation. Unless, of course, that pagefile is full in which
case Windows will use a different pagefile rather than produce an
error message.

The only actions that are associated with a specific pagefile are the
system failure memory dumps, which will always go to the pagefile on
the system drive.


Ron Martell Duncan B.C. Canada
--
Microsoft MVP
On-Line Help Computer Service
http://onlinehelp.bc.ca

"The reason computer chips are so small is computers don't eat much."