Matthias Scheler <tron%zhadum.org.uk@localhost> wrote:
What would be the sense in keeping allocated pages in memory even if
they haven't accessed for hours? Using the memory e.g. as disk
buffers will reduce the total amount of disk I/O which is what
a modern VM system tries to achieve.
That may be true for a desktop system, but NEVER for a server system.
If lighttpd is swapped because it wasn't used for some time, that's
bad. If the python scripts I use via fcgi are swapped out (there are 4
threads of them, so it's very likely one will get swapped out after
some time because not all are always used - and which one is used is
randomly), it's even worse. So the probability is high that the user has
to wait about 10 seconds till he gets a reply from for example hg
(remember, first lighttpd needs to beloaded from swap, then one of the
python threads). Is this desired behaviour for a server? Nope, it's
really not, definitely not. This really kills the whole idea of having 4
threads so that no new thread needs to be spawned, which would take
too long for the user to wait. Same for spamd which runs with 5 threads
here. Kills really the point of having multiple threads to speed up
spawning. It kills the whole idea of pre-spawning.
Please use Google to look for "Using swap is bad for performance"
and you will find some interesting comments on the topic.
On a desktop maybe, but not on a server. And as far as I know, NetBSD
is used more for servers than for desktops, though it's a very nice
system for a desktop as well.