Please be careful with your quoting, for someone not new to this thread
it would be very difficult to tell what you wrote and what I wrote. The
problem is your mail client. Almost any other client on the planet can
do a better job than the one you are using.
On Mon, 4 Aug 2003 21:34:45 +1000
"Ben de Luca" <[EMAIL PROTECTED]> wrote:
> I wasn't really attacking your post more the article, It was flawed in
> so many ways its not funny.
I was commenting about the obvious outrage in your post.
> Its not that I just feel that, I and you both know they are flawed.
If the advantages of something like Gentoo was as obvious as Gentoo's
proponents suggest, I would expect them to jump out and grab me by
the throat. They didn't.
> So I am offering up my time to run the tests at work.
>
> What things are suggested as to what should be done.
>
> I think the test should use the same hardware at least, I will run it
> multiple times.
As I suggested, this test dones't require much in the way of resources.
In fact, the test could easily be done on the same machine. I would even
suggest that they be done on the same partitions. Ie:
0) Install distro #1
1) Test distro #1
2) Blow away distro #1
3) Install distro #2
and so on.
> That for any files that are opened they should be on a read only part
> ion that is not the one the install is on, hence negating the effects of
> disk on the test.
If they are running on the same disk on the same partition (at different
times) the disk can be discounted.
> That the tests should be repeated enough times? Any one know how to do a
> Power analysis so I know we will get useful results?
The ideal benchmark to run on this is something like nbench:
http://www.tux.org/~mayer/linux/bmark.html
Other tests would include a kernel compile (using the same .config file).
Tests like timing open time of some_app are mostly useless.
> I am interested in the gentoo what optimizations flags would be suggest?
> Just 03?
Ask the Gentoo people.
> What distro's will race? Should I turn of systems services? I would hate
> a cron to skew the results? Or should I just leave the default service
> settings?
Regardless of what the setup is, it should be the same across all distros.
> I guess I should provide details of any changes I make? So the tests can
> be compared?
Yes, yes.
> Should I use different hardware?
That only makes it more difficult. Pick one current pretty much standard
machine.
> What specific tests should I run? I would like to test, for things other
> than just opening a spread sheet?
OPenning is speadsheet is a particularly stupid test.
> How about
>
> Hitting a db?
Too hard.
> X rendering?
Possibly, but I don't know of any good benchmarking program.
> File access?
Too artificial.
> Compiling a kernel?
That is a good test because a developer does a lot of compiling.
By do look at nnbench (above).
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo [EMAIL PROTECTED] (Yes it's valid)
+-----------------------------------------------------------+
"The RIAA is obsessed to the point of comedy with the frustration
of having its rules broken, without considering whether such rules
might be standing in the way of increased revenues. Indeed,
Napster and Gnutella may turn out to be the two best music-marketing
gimmicks yet devised, if only the RIAA would take its head out of
its ass long enough to realise it."
-- Thomas C Greene on www.theregister.co.uk
--
SLUG - Sydney Linux User's Group - http://slug.org.au/
More Info: http://lists.slug.org.au/listinfo/slug
