On 29 Nov 2002, Perry E. Metzger wrote:
> Brian Chase <vaxzilla@jarai.org> writes:

> > I was astonished with how close the results were, and obviously a bit
> > disappointed to not see any gains.  Maybe there are none to be had
> > with my whopping 12MB of available memory? :-)
>
> It hardly seems surprising at all. I doubt you could get to 95% use by
> anon on a 12M machine -- where would files and program text end up?
> On a larger memory machine, I doubt it would improve performance
> either. anonmax of 95 seems very silly to me actually...

So somewhere between Simon Burge's 8MB pc532 and my 16MB VAX, there's a
zone where increasing anonmax can be useful.  I really don't have a very
good idea of what anonmax does yet, but based on Simon's experience,
I thought it'd be worthwhile to see if changed anything on my VAX.
My astonishment wasn't so attributed to the lack of improvement in
performance; rather it was that, after 5 hrs of compiling, the build
times came within 4 seconds of eachother.  I've a feeling they'd have
been even closer had I not poked at the keyboard a bit during the
first build.

One thing I think would be useful for NetBSD is some document that's
equivalent to Sun's _Solaris Tunable Parameters Reference Manual_
<http://docs.sun.com/db/doc/806-7009>.  A general performance tuning
guide would be useful, too.  At least it would be for those who don't
have an  intimate relation with the kernel source, but who could still
benefit from a technical overview of the kernel geared towards
sysadmins.

-brian.