> I've heard that there's some patch into kernel which makes swapping
> over nfs to actually WORK.
Do you got an URL to it?
After hearing about some theoretical figures that said it would be a
performance increase swapping to the net (compared with swapping to local
disc) when you have a couple of boxes in a local net (even on 10MBit/s) I
thought about implementing this (if possible due to the deadlock problems
you mention) in two user-space processes, one to swap out data on the net
and one to store other machines swapped out memory (it will probably
require some kernel-hacking, but I would try my hardest to keep
stuff like that to a minimum).
Splitting it in two separated processes would make it system independent,
so you could run the swap-client on some Big Box with lots of memory
(even though the theoretical figures said that you gain the most when the
network has lots of equally configured boxes, memorywise). It would also
be possible the implement the server into something else than linux.
It probably hasn't much real-life application (I would never trust even a
local net with parts of my core) right now, but as 100MBit/s networking
become more and more popular it could be fun to experiment with.
Anyone who has looked at the swapping part of the kernel and know if my
idea is even doable?
I would apreciate any comments and ideas -- I've just recently begun
thinking about this and could possibly be fundamentally wrong about
something.