[3517] in RedHat Linux List

home help back first fref pref prev next nref lref last post

Re: UPS reccomendations

daemon@ATHENA.MIT.EDU (Allen Francom)
Fri Nov 8 21:25:41 1996

Date: Sat, 9 Nov 1996 09:28:24 -0800 (PST)
From: Allen Francom <afrancom@numedics.transport.com>
To: redhat-list@redhat.com
cc: Mac McClellan <list@digitalpla.net>, redhat-list@redhat.com
In-Reply-To: <19961109023311.29294.qmail@brouhaha.com>
Resent-From: redhat-list@redhat.com
Reply-To: redhat-list@redhat.com



Nice Job Eric !


This should go into the UPS-HOWTO !  - Seriously


BTW - The smaller APC units still output square waves.

      It is interesting to note that when reading the current with 
      an inexpensive digital meter, it will not show as many volts
      as it says it should.

      For example, I was scared that my 120v APC little guy for one
      computer was only putting out about 104v, until I used 
      my electrical engineer friend's beautiful analog meter, which
      somehow knew to show 118v+/-.

   Hmn.  So much for digital technology...

THX
-AEF


On 9 Nov 1996, Eric Smith wrote:

> Mac McClellan <list@digitalpla.net>
> > I have three servers that I want to put on a UPS.  Each server has a 300W
> > power supply.
> 
> UPSes are generally rated for their maximum load in VA rather than Watts.
> If you make the simplifying assumption that the power supply draws constant
> current, then VA = 1.4 * W, so a 300 W load would be 420 VA.
> 
> However, the 300 W rating of your power supply is a maximum, and the actual
> power consumption will vary depending on what is installed in the machine.
> 
> Another thing to consider is that while your UPS must be rated for at least
> the VA of your load, there are reasons why you might want a higher rating:
> 
> 1)  A typical UPS is designed to power the rated load for about 5 to 10
>     minutes.  A unit with a higher rated load will power the lesser load
>     for a longer period of time.  In fact, lead-acid batteries are more
>     efficient if you discharge them at less than their peak current, so if
>     you were to compare two UPSes rated at 500 VA and 1000 VA that both could
>     power their rated load for 5 minutes, the 1000 VA UPS might be able to
>     power a 500 VA load for noticably more than 10 minutes.
> 
> 2)  Lead-acid batteries (even the "deep cycle" ones) degrade when they are
>     cycled.  If you need to power the system for 5 minutes, you are better
>     off getting a UPS that is capable of powering the system longer, but
>     setting up the UPS or the software to turn it off sooner.  That way the
>     battery doesn't get discharged as much, and will have a longer lifetime.
> 
> > What kind of UPS would be the best for my situation?
> 
> There are three main kinds of UPSes:
> 
> 1)  Standby Power Supplies:  These have a relay to switch from the AC line
>     to the inverter (battery power) in an outage.  They have a rated transfer
>     time in milliseconds, which should be as low as possible.
> 
> 2)  Online UPSes:  These always run the load from the inverter, even when
>     AC power is available.  The advantage is that there is no transfer time,
>     and that this provides voltage regulation to the load.  These tend to
>     be very pricey.
> 
> 3)  Line Interactive UPSes:  Despite reading a technical note purporting to
>     explain how Line Interactive UPSes work, I still don't really understand
>     it.  These seem to combine the characteristics of the other two types.
>     APC claims that they are more reliable than other types.
> 
> Another thing to consider is the output waveform.
> 
> 1)  Cheap UPSes have a square wave output.  Switching power supplies in
>     computers are perfectly happy with this.  But other devices such as
>     monitors and printers, and any devices using power transformers, may
>     not be so forgiving.  Transformers are substantially less efficient on
>     square wave input, so they use more power (draining the UPS battery
>     faster), and turn that extra power into heat.
> 
>     I'm not sure whether anyone is still selling these.
> 
> 2)  Slightly more expensive units have modified square wave output.  These
>     try to round the edges of the square wave a bit.
> 
> 3)  Quality units have true sine wave output.
> 
> The relevant measure of the output waveform is harmonic distortion; the
> lower the distortion the better.
> 
> And the other main distinguishing feature is the computer interface.
> 
> 1)  No computer interface at all.  Don't use this unless you are certain that
>     you only have to ride through brief power interruptions.
> 
>     I'm not sure anyone sells these any more.
> 
> 2)  Simple signalling.  These have one or more sets of relay contacts which
>     may be used  to signal the computer.  One set of contacts notifies the
>     computer that the AC power has failed.  Typically another set indicates a
>     battery low condition.  There may or may not be a way for the computer to
>     signal the UPS to shut it down.  The init and powerd programs from the
>     SysVinit package (standard on Red Hat Linux) include support for simple
>     signalling.
> 
> 3)  Smart.  These have an EIA-232 serial interface or a network interface for
>     SNMP.  Usually they support dumb signalling also.  Unfortunately some of
>     the major manufacturers try to keep their serial protocols secret to
>     force you to buy their expensive software (of which they have none for
>     Linux anyhow).  Fortunately people have reverse engineered enough of
>     the protocol for the APC SmartUPS to make it work with Linux; there is
>     a package called apcd.  I'm not sure whether there is an RPM for it.
> 
> > Should I get one big one for all three or three individual smaller ones?
> 
> There is some economy of scale to using one large unit instead of three
> small ones.  You can have one machine talk to the UPS, and notify the other
> two via your network when they should shut down.
> 
> I personally prefer using a smaller UPS for each machine, with enough
> excess VA rating that if one of my UPSes fails, I can plug a machine into
> another machine's UPS while I get the broken UPS fixed.
> 
> > What about software to shutdown the servers when they run out of juice?
> 
> See the section on signalling above.  I have my machines configured to
> shut down and turn off the UPS after five minutes of AC outage.  I bought
> UPSes with sufficient rating and battery capacity to run my machines for more
> than 30 minutes, so that won't cycle the batteries too much.
> 
> Cheers,
> Eric
> 
> 
> --
>   PLEASE read the Red Hat FAQ, Tips, Errata and the MAILING LIST ARCHIVES!
>   ________________________________________________________________________
>   http://www.redhat.com/RedHat-FAQ   http://www.redhat.com/RedHat-Errata
>   http://www.redhat.com/RedHat-Tips  http://www.redhat.com/mailing-lists
>   ------------------------------------------------------------------------
> To unsubscribe: mail -s unsubscribe redhat-list-request@redhat.com < /dev/null
> 


--
  PLEASE read the Red Hat FAQ, Tips, Errata and the MAILING LIST ARCHIVES!
  ________________________________________________________________________
  http://www.redhat.com/RedHat-FAQ   http://www.redhat.com/RedHat-Errata
  http://www.redhat.com/RedHat-Tips  http://www.redhat.com/mailing-lists
  ------------------------------------------------------------------------
To unsubscribe: mail -s unsubscribe redhat-list-request@redhat.com < /dev/null


home help back first fref pref prev next nref lref last post