[Hampshire] Server swappiness

Hi all,

We have a Linux server that mostly just runs VMWare Server hosting guest
VMs
for development and Samba. An issue we often see is that VMs will tend to
get
swapped out, making them incredibly slow to access after some period of
inactivity.
The server has 16GB of RAM, which is plenty for the VMs it is running.

I'm guessing that what is happening is that the VMs are being swapped out
to make
way for file cache. An rsync backup gets run every night which probably
doesn't
help things in this regard.

What we'd like to do is to configure things so that filecache doesn't grow
at
the expense of VMs. Some reading around has highlighted the vm.swappiness
parameter (or /proc/sys/vm/swappiness) which seems to control what I'm
interested
in, however I haven't found any good description of what sort of effect
different
values have.

Currently, the server has a swappiness of 60.

Some suggestions have recommended setting it to 0.

Can any explain, or have pointers to documentation, which give examples of
what
sort of behaviour to expect for different values?

I.e., does 0 completley switch off filecache, or simply stop it swapping
stuff
out? Does a value of 0 cause problems?

How about 10, or 30? I've not idea whether dropping it to 30 would have any
affect at all, a little affect or lots of affect. i.e., given that we run
an rsync over 113GB of files, will anything other than 0 make any
difference?

Thanks.

--
Samuel Penn
sam@???