Need ball park power and cooling requirements
Ben Ransom
bransom at ucdavis.edu
Tue Oct 1 14:58:07 PDT 2002
At 03:34 PM 10/1/02 -0400, you wrote:
>On Tue, Oct 01, 2002 at 10:09:52AM -0700, Ben Ransom's all...
> > At 04:01 PM 9/30/02 -0600, Dean Johnson wrote:
> > >On Mon, 2002-09-30 at 15:17, Timothy H. Keitt wrote:
> > > > I'm setting up lab space and plan to build a small (8-16 node) cluster
> > > > of dual Xeon nodes (Dell 2650 or equivalent). I need to give the
> > > > engineers the power and cooling needed in the 6'x8' server room that
> > > > will house the cluster. Any rules-of-thumb for what's required? Thanks.
> > > >
> > >
> > >A good ballpark, with decent margins, would be 300w per box. You need to
> > >add things like monitors and switches too.
> > >
> > > -Dean
> >
> > I'd add a little more margin. I'm using dual athlons, which may be more
> > power thirsty than Xeons -- dunno. Nonetheless, planning on just the power
> > supply rating per box could be limiting. Our cluster sometimes goes over
> > calculated amp (and power) ratings when all cpus are crunching 100%. We
> > have Baytech power strips rated at 20 amps which should have handled 7
> dual
> > athlon boxes apiece, if going by the spec numbers. We had to add another
> > power strip and spread out to only 5 boxes each. Fortunately, I had
> enf 20
> > amp circuits in the room (4 for a 20 node cluster + eth switch + head
> node).
>
>3.5 amps per machine? MPs on what board? how much ram and how many HDs?
>
>/kc
dual Athlon 1900MPs on Tyan 2466N, 1g Ram, 1 60gb atadrive each (pvfs), and
wulfkit.
Ack, perhaps I misled things just a little in saying 5 boxes each. I could
do 6 each circuit, but the point was that 3 power strips was not enf for
all 20 nodes, so I now spread the 20 nodes evenly across 4 20amp
strips. The Baytech power strips report peak amps. I got readings of 19
amps with 6 nodes plugged in and running 100%cpu. I got over 21 amps with
7 nodes plugged in, 100%cpu. On this, the Baytechs didn't trip until over
22amps, so I guess they throw in a little fudge factor in our favor. On
the other hand, the Baytechs were reporting higher load than should be
possible from the power supplies. I think these are PCF power supplies,
but that is fuzzy memory now. I think they are 300w which should be max of
~2.75amps. So, 6 boxes * 2.75 should be max draw of 16.4 amps, yet the
Baytechs reported 19amp at 100% cpu utilization.
-Ben
> >
> >
> > -Ben Ransom
> > Programmer/ Systems Administrator
> > Mechanical & Aeronautical Engineering, UC Davis
> > (530) 752-1834 Fax: 752-4158
> >
> > _______________________________________________
> > Beowulf mailing list, Beowulf at beowulf.org
> > To change your subscription (digest mode or unsubscribe) visit
> http://www.beowulf.org/mailman/listinfo/beowulf
>
>--
>Ken Chase, math at velocet.ca * Velocet Communications Inc. * Toronto,
>CANADA
More information about the Beowulf
mailing list