I've found the cause for vhand eating too much memory. Well, I think
I have... I think it is because kswapd would wake up vhand after 3
failures, while vhand would guarantee that 1/32nd of memory is ready
to swap out... Now, putting KSWAPD_MAX_FAIL at 128 (!) has alleviated
this problem, it's not over yet, but at least it's not severe.
(at the moment I'm looking at an uncharged kswapd and 5% max for
vhand. probably vhand gets billed for kswapd :-)
... somebody with an Alpha (HZ=1024) should take a closer look at
this...
I also merged my patch with the excellent network_atomic_alloc_59
patch from Zlato Calusic <Zlato.Calusic@CARNet.hr>. Thanks Zlato!
-------------
Vhand is a patch that improves system responsiveness when memory
is short. It does this by also aging shared memory and buffers.
It also applies a fairer aging sceme to the aging of normal
program pages.
Although this patch is reported to be rock-solid, I still consider
it Beta. This is mainly because I've only had about 15 success
stories in my mailbox (and no failures:-).
You can find the patch on linux-mama <http://www.huwig.de/linux/mama/>
and on my home page <http://www.fys.ruu.nl/~riel/>.
-------------
Please send success-stories, bug-reports, bugfixes and flames
to me. Anything will help...
Rik.
----------
Send Linux memory-management wishes to me: I'm currently looking
for something to hack...