Re: Linux 2.4.4-ac10

Mike Galbraith (mikeg@wen-online.de)
Sat, 19 May 2001 08:45:57 +0200 (CEST)


On Fri, 18 May 2001, Rik van Riel wrote:

> On Fri, 18 May 2001, Stephen C. Tweedie wrote:
> > On Fri, May 18, 2001 at 07:44:39PM -0300, Rik van Riel wrote:
> >
> > > This is the core of why we cannot (IMHO) have a discussion
> > > of whether a patch introducing new VM tunables can go in:
> > > there is no clear overview of exactly what would need to be
> > > tunable and how it would help.
> >
> > It's worse than that. The workload on most typical systems is not
> > static. The VM *must* be able to cope with dynamic workloads. You
> > might twiddle all the knobs on your system to make your database run
> > faster, but end up in such a situation that the next time a mail flood
> > arrives for sendmail, the whole box locks up because the VM can no
> > longer adapt.
>
> That's another problem, indeed ;)
>
> Ingo, Mike, please keep this in mind when designing
> tunables or deciding which test you want to run today
> in order to look how the VM is performing.

I've bent your code up a bit. I've not yet been tempted to replace
any of it with a knob ;-) There is a little piece I'd like to see
thrown away though.. the loop in refill_inactive does nothing good.

The test I prefer is a good one for the area of vm performance I'm
most interested in. It doesn't cover the full vm spectrum by any
means. I don't have a setup (any) good for testing mondo network or
IO stuff. I test a simple 'job one size to large' scenario. Yes,
it's limited test coverage.. it's still legitimate.

Perhaps when you're evaluating vm performance, you should try my
simple test once in a while. :) I'll bet you a bogobeer right here
and now that when 2.4.5 hits the street you're going to be queried
by the big-busy-box folks wrt swap volume.

> Basic rule for VM: once you start swapping, you cannot
> win; All you can do is make sure no situation loses
> really badly and most situations perform reasonably.

I disagree with that. I've seen a heavily swapping box run like
a scaulded ass ape many times.

Warsteiner,

-Mike

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/