[Beowulf] UPS system for Linux cluster

Lux, James P james.p.lux at jpl.nasa.gov
Thu Apr 30 11:57:55 PDT 2009

> -----Original Message-----
> From: Joe Landman [mailto:landman at scalableinformatics.com] 
> Sent: Thursday, April 30, 2009 11:35 AM
> To: Lux, James P
> Cc: Tom Pierce; vlad at geociencias.unam.mx; beowulf at beowulf.org
> Subject: Re: [Beowulf] UPS system for Linux cluster
> Lux, James P wrote:
> > 
> > Actually, this isn't particularly surprising to folks who deal with 
> > electrical power. It *is* surprising to folks who see 
> volts*amps (VA) 
> > and think it means watts.
> > 
> > Most power conversion equipment has a higher VA rating than Watt 
> > rating (and, so do most loads, so it works out ok in practice)  PF 
> > (Power Factor) = Watts/VA, and a typical PF is around 0.85.
> Most of the advice we've been given about "arbitrary" gear 
> from power folks has been to use 0.7 (approximately 
> 1/sqrt(2)) as your PF, to account for varying supply quality 
> and leave an engineering margin (for error).  Indeed, looking 
> at some of the PF curves for supplies under heavy load shows 
> PF for "85%" supplies dipping into the mid to upper 70s.

I think a power supply that bad would have trouble passing the usual regulatory standards (IEC, UL), which put limits on the power factor. 

See, e.g., EN61000-3-2 (which became mandatory on 1 Jan 2001)

"80 plus" requires PF>0.9

More information about the Beowulf mailing list