[_] Sun X4100 performance (was: Sun X2100 - for a webserver)
Jan Grant
jan.grant at bristol.ac.uk
Fri Nov 3 13:26:59 GMT 2006
On Fri, 3 Nov 2006, Iain Hallam wrote:
> Jan Grant wrote:
> > On Wed, 1 Nov 2006, Iain Hallam wrote:
> >
> >> I have just been running extensive tests on identical X4100s with 12 GB
> >> RAM and dual CPUs for number crunching. We essentially pitted Red Hat
> >> Enterprise AS against Solaris 10 and discovered that the Red Hat box had
> >> a performance advantage of 3:2
> >
> > That sounds so spectacular that it makes me think something's broken.
> > The OS shouldn't get in the way of straight-line computation that much -
> > what were the benchmarks, out of interest?
>
> It was a series of tests based on the every-day usage of our stats
> package from researchers. The computation part was probably pretty even,
> but there's a fair amount of writing intermediate results files out for
> further analysis, and I suspect from the numbers that it's the disk
> systems that are causing the difference.
Be interested to know if you get the same results using (say) ext2 and
noasync.
> One other factor in this particular scenario is that our researchers are
> heavily biased to a single package which is developed on Linux and
> optimised for it. (Hence the warning about application variance.) That
> is what is likely causing the slower usage of the hard disk in the above
> results.
OK, that's interesting, cheers.
--
jan grant, ISYS, University of Bristol. http://www.bris.ac.uk/
Tel +44 (0)117 3317661 http://ioctl.org/jan/
Goth isn't dead, it's just lying very still and sucking its cheeks in.
> Jan Grant wrote:
> > On Wed, 1 Nov 2006, Iain Hallam wrote:
> >
> >> I have just been running extensive tests on identical X4100s with 12 GB
> >> RAM and dual CPUs for number crunching. We essentially pitted Red Hat
> >> Enterprise AS against Solaris 10 and discovered that the Red Hat box had
> >> a performance advantage of 3:2
> >
> > That sounds so spectacular that it makes me think something's broken.
> > The OS shouldn't get in the way of straight-line computation that much -
> > what were the benchmarks, out of interest?
>
> It was a series of tests based on the every-day usage of our stats
> package from researchers. The computation part was probably pretty even,
> but there's a fair amount of writing intermediate results files out for
> further analysis, and I suspect from the numbers that it's the disk
> systems that are causing the difference.
Be interested to know if you get the same results using (say) ext2 and
noasync.
> One other factor in this particular scenario is that our researchers are
> heavily biased to a single package which is developed on Linux and
> optimised for it. (Hence the warning about application variance.) That
> is what is likely causing the slower usage of the hard disk in the above
> results.
OK, that's interesting, cheers.
--
jan grant, ISYS, University of Bristol. http://www.bris.ac.uk/
Tel +44 (0)117 3317661 http://ioctl.org/jan/
Goth isn't dead, it's just lying very still and sucking its cheeks in.