[MLton] Paging under RedHat 9.0 although largely fitting in core (resend)
Henry Cejtin
henry@sourcelight.com
Wed, 10 Mar 2004 17:28:16 -0600
Like I said before, I remember seeing what the problem could be, where the
kernel would rather cause a process to swap than purge file buffers/cache
from memory, but I thought that that was pretty much fixed up these days.
If it is the problem that I recall, it could be `fixed' with the following
hack program: it creates a file, writes a ton to it, then closes it and then
removes it. This was an attempt to free up blocks that were being used for
the buffer cache.
Can you tell if you are actually swapping as opposed to kswapd simply burning
CPU time? You can tell the difference by running procinfo, sleeping for 60
seconds and then running it again. Look at the differences in user, nice,
system and idle to see how much the CPU was used in between, and look at the
difference in page in/out and swap in/out to see if any disk activity is
going on.