Re: Swap performance statistics in 2.6 -- which /proc file has it?

[Stephen Satchell <list@satchell.net>]

On Tue, 2003-12-09 at 11:53, Dominik Kubla wrote:
> Richard B. Johnson wrote:
> > If you need statistics v.s. time, you need to write an application
> > that samples things at some fixed interval. In a previous life,
> > I requested that "nr_free_pages()" be accessible from user-space,
> > probably via /proc. That's all you need. Maybe that could be
> > added now?  In any event, samping free pages at some fixed-time
> > interval should give you all the information you need.
> 
> vmstat -a
> sar -B
> sar -r
> 
> O'Reilly's "System Performance Tuning" might make for an interesting read,
> especially pages 110ff (also its Linux informations are a bit out of date).

How does sampling free pages give you an accurate measurement of swap
activity?  If I look at the free-page count at one-minute intervals, the
system can, and WILL, inhale and exhale pages at a frightening clip, and
there is no way I can see that sampling free-page count in a
low-overhead way will do the trick.

How does vmstat disk-swap activity?  Looking at the source for vmstat in
procps-2.0.11 I see how they do it for 2.4 kernels, but the part for 2.5
kernels doesn't seem to try to pick up swap statistics at all -- because
there are none to get?

(signed) Puzzled.

(Why does this whole discussion remind me of the Firesign Theatre album
_I Think We're All Bozos On This Bus_, and the question that Ah Clem
asked:  "Why does the porridge-bird lay his eggs in the air?")