On 5 mrt, 14:06, "Mark A" <no... (AT) nowhere (DOT) com> wrote:
I don't know how the Linux swapping algorithm works, but I meant to
say that putting something on disk is not necessarily a bad thing. The
swapping itself can cause performance problems, but having something
statically in swap does not. It even clears up space that can be used
for something more useful. In that regard, as long as there's no
(excessive) swapping occurring, I wouldn't bother about Linux using
some swap space, it could be perfectly rational. You could even
compare it to DB2 flushing dirty pages.

But then again, I'm no Linux expert. I'll check my own systems on
Monday.

--
Frederik Engelen

On Mar 5, 8:25*pm, Frederik Engelen <engelenfrede... (AT) gmail (DOT) com> wrote:
Mark,

I just ran this for about 15 minutes:
[*@* ~]$ free;date; vmstat 2 | awk '$7!~/^0$/||$8!~/^0$/{print
$0}';date
total used free shared buffers
cached
Mem: 32963300 32744332 218968 0 313136
19753432
-/+ buffers/cache: 12677764 20285536
Swap: 71681956 553488 71128468
Mon Mar 7 10:53:02 CET 2011
procs -----------memory---------- ---swap-- -----io---- --system--
-----cpu------
r b swpd free buff cache si so bi bo in cs us
sy id wa st

Mon Mar 7 11:07:19 CET 2011

There's 550MB of swap used, but nothing is retrieved or written out,
so there should be no impact.

Monday morning might not be my most productive moment, but this is a
representative sample. IO measurements not shown, but values for bi
vary around a few thousand K/measurement. Server hosts 30 databases
and the most active one will have processed 200K transactions during
this time and executed a bit over 2M statements.

I'm curious what you (and Helmut) think of these numbers.

--
Frederik Engelen