Re: Page file size?

The message <jm847mCeXJGIFAtT@xxxxxxxxxxxxxxx>
from GSV Three Minds in a Can <GSV@xxxxxxxxxxxxxxxx> contains these words:

Bitstring <5b2h14lokn00rlrams14q41ujdr2bl58ju@xxxxxxx>, from the
wonderful person Terry Pinnell <terrypin@xxxxxxxxxxxxxx> said
"Rob" <noone@xxxxxxxxxxxxxxxxx> wrote:

"Terry Pinnell" <terrypin@xxxxxxxxxxxxxx> wrote in message
news:qehg14tcb9o42t46l7fn1744otc60hau7i@xxxxxxxxxx
What do the experts here think about the optimum pagefile size and
location for my new XP Pro PC please. It has 4 GB RAM and two
(nominally)750 MB HDs, neither partitioned as yet. The default size
set is 2046 MB 'Total paging size for all drives' and is on C: (the OS
drive).

I've tried on many systems, and in general, a size of 1.5 to 2x
the installed RAM works nicely. If you can put it on a drive
other than the system one, even better.

HTH,

Ah, thanks Rob, that seems to crystallise my earlier view that I
should put it on I: (as I did on my old PC). However seems at odds
with Dean's point about a pagefile being needed on C:?

The whole point with pagefiles is that you really don't want the system
to be using them anyway (unless you are hot swapping among users or
something), so where you put them is not very exciting. Yeah yeah, they
run somewhat faster on a different physical drive from the main OS, but
they are still three orders of magnitude slower than RAM, so you really
don't want them used except as a desperate last resort.

I agree, but, it's still worth considering how to optimise them to stop
the performance hit extending to _four_ orders of magnitude on those
hopefully rare occasions that the system actually accesses them. :-)


You want to have SOME, since otherwise Windows assigns space in real RAM
to imaginary areas of data and program which don't actually exist. It
can assign them to a paging file just as well (if one exists) and since
they are imaginary they don't even need reading/writing (although the
pagefile will claim to be assigned). There are utilities to allow you to
watch this actually (not) happen, at least in Win2k/XP.

Yes indeed, the whole process of speculative demands by programs on the
OS's resources start to look like a Monty Python sketch along the lines
of "Albatros on a stick". :-)

Having some on the system drive avoids windows getting in a knot if the
assigned paging drive is offlined.

Yes been there etc. Mind you, it's not entirely insurmountable, just a
bit of faffing about if I remember the last (and only) time I lost
access to the drive containing the pagefile. On that occasion, I think
the system forced an increase from the 2MB one I'd minimised it to and I
was able to resize it for the duration of the drive outage.

However, I've just noticed that I've created dynamic pagefiles on all
the other disk volumes with 2MB minimums and 1536MB maximums (that's
aside from the permanent 1.5GB one in the first partition of the second
drive and the permanent 2MB one in the OS partition). This should
mitigate against any further 'inaccesable' pagefile messages in the
unlikely event of a repeat second disk failure.

Looking at the actual sizes of these additional pagefiles suggests the
OS has a preference for using the fixed size pagefile over the
dynamically sized ones, so it seems a useful 'backup' strategy with a
minimum of performance loss.

The 'lets use 3x the real RAM' is a heuristic derived from what was a
sensible amount of VM to assign to a multitasking time share system
(Multics, IIRC) and has no validity in the real world. Out here, the
right answer is 'whatever your maximum memory usage might be, less the
amount of real RAM you can afford', plus a half-gig or so to take care
of the 'imaginary paged out areas' mentioned earlier.

Very sound advice with which I totally concur.

--
Regards, John.

Please remove the "ohggcyht" before replying.
The address has been munged to reject Spam-bots.

.