Irrelevant to whether that use of the page file that he
can see is BEING DONE USING FREE RESOURCES.
If that's "free resources" then we can as easily argue that
every single thing windows or an app is doing is merely
using "free resources". Use of a bit of CPU time, memory
access, drive controller and drive. Just like having two
apps trying to run alongside each other. You can't *defer*
virtual memory, the whole point of it is that it's actively
happening as needed, but the "need" is a dumb logic that is
too conservative for well endowed systems, depending on how
the amount of physical memory compares to the requirements
of the jobs that system runs. To be most efficient there
would need be a user-selectable or learning profile and a
use pattern that remains fairly static - something we can't
necessarily assume on a PC.
If you're trying to say that windows waits till the system
utilization is low, that's about the last time there was a
need to free up memory, but by paging anything out it
creates the opportunity for the next use session to require
that paged data read back again when it was most important
for it to not be paged out. It can help to page out unused
data ahead of time, but it is done too aggressively within
the assumption it is a good thing to do when it is too often
just a wasted effort, and as mentioned above, then has to be
read back into physical memory because it can't predict the
future without a crystal ball.
What really happens is that applications allocate more
memory than they use, and the pagefile utilization figures
reflect this allocated (but often remaining unused) virtual
memory space. It is not likely that John's system (as an
example since he reported a figure) is actually paging out
most of that 500MB as data rather than just reserved,
allocated virtual space, but nevertheless there is pagefile
access ongoing, it is still a minor performance hit relative
to not having it happen. The question is really whether the
penalty is worse that the potential that having no paging
allocation would result in out of memory errors, something
that depends entirely on the actual uses of the system.
Some people can turn off their swapfile and run like that,
and others will do something mundane like loading a game,
only to find an error generated about 15 minutes into
gameplay (which I have done, the system had ample free
physical memory but windows wanted to allocate more and
doesn't like not being able to do what it wants, regardless
of lacking a real need).
It is slower than not doing it, as it does happen during
NON-idle times, right in the middle of heavy use. It
certainly did when it generated the errors during the
gaming. After investigation I did re-enable the pagefile on
the system exhibiting the problem to confirm this account.
