> You can think of it this way. The kernel has two free-page lists:
> one list of pages with nothing interesting in them, and another of
> pages containing stuff that might become interesting and could be
> just handed over without any I/O. Looking at it this way, you
> want the second list to be large and the first to be small. And
> that's what the kernel tries to give you.
However, in the latest kernels, the behavior of the cache is unacceptable on
low-memory machines such as mine. The kernel actively caches WAY too much,
preferring cache over real programs in main memory. This, I believe, is very
much broken, and the performance of pre9 is far worse than 2.0 kernels in
that respect. People with a lot of memory probably don't feel this at all.
Symptoms? While reading mail in mutt, and reading the disk in some other
process, mutt just might be swapped out of memory, requiring a few seconds
to come back to life when I want to read the next message. That isn't the
Linux I'm used to...
Is there any chance that this will be fixed before 2.2.0, even if the fix
only involves bringing 'pagecache' and 'buffermem' back to life? I hope.
Nimrod
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/