[Beowulf] powering up 18 motherboards

Robert G. Brown rgb at phy.duke.edu
Fri Feb 18 05:11:17 PST 2005

On Fri, 18 Feb 2005, Alpay Kasal wrote:

> James Lux, thanks for the extremely useful explanation. Btw, I'm in
> Brooklyn, NY. 120volts, 60cycles, regular AC power. I don't know the gauge
> of the wiring in the walls but (as mentioned in another response just now) I
> suspect it is old wiring and is the reason for the strange 10amp circuit
> breakers.
> I looked at the x10 modules. Seems like it could be very useful, just script
> all of them from my headend. For now I'm going to try to handle the power-on
> sequence myself. I figured I could steal 3 10amp circuits from the house.
> Follow me... Turn on 4 nodes (on 1 strip) which will peak at 5.2amps. let
> that settle down to a steady 3.48amps and hit another strip of 4 nodes.
> Total draw while the 2nd batch is starting is 8.68amps. It should steady at
> 6.96. I then have room to turn 1 more node on. Then one more after that. A 4
> step process to get 10 nodes powered up without going over 10amps. Perform
> the same exact steps on a 2nd circuit. Annoying but possible without
> spending anymore money.

You problem will occur when the power goes off and comes back on when
you aren't there.  We have rather frequent 5-10 second powerouts down
here -- without UPS's I used to go nuts in my house.

> I was really hoping a decent $200-300 UPS would come to the rescue here. Oh
> well.

I don't think that putting one of these on per circuit is a bad idea;
the real problem is that a UPS might draw more than your lines' capacity
when initially charging -- I don't know for sure how much of a load the
divert to charging when passing a load through.  If you have >>a<<
bigger circuit, you might be able to charge one fully on it, move it,
plug everything in and power everything up, and use it as a line buffer
of sorts.  Even a couple of very cheap $50 ones that only will give you
a minute or two might keep you from blowing the CBs every time the power
in your area bobbles.  Assuming that it does bobble -- maybe NYC never
has power issues, even when dogs piss on transformers...;-)

> I just had a thought... I planned on making use of wake-on-lan. I can just
> start sending jobs to the whole network though if all of it is asleep, I'd
> have to still be careful of the powerup-sequence. Grrrr. Maybe a script to
> perform WOL before starting any number crunching.

Yeah, that's an alternative.  Leave one box set to power up, set the
rest to NOT power up after a power outage, if you can, and power them up
with WOL from a script.  But yes, a PITA.


> Boy did I take nice big fat electrical lines for granted in the past!
> Alpay
> -----Original Message-----
> From: Jim Lux [mailto:James.P.Lux at jpl.nasa.gov] 
> Sent: Thursday, February 17, 2005 7:18 PM
> To: Dean Johnson; Alpay Kasal
> Cc: beowulf at beowulf.org
> Subject: Re: [SPAM] [Beowulf] powering up 18 motherboards
> No, the UPS won't help.  It might make things worse, because as you flip on 
> all that load, the voltage will sag, causing the UPS to turn on, which then 
> might trip from the overcurrent (assuming you're not out buying a 2kW UPS).
> You could use the X-10 type (aka Plug n Power) remote controlled relays 
> (don't use Lamp modules.. you need Appliance modules, which are relays
> inside).
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu

More information about the Beowulf mailing list