[Beowulf] Efficient UPS Aids Google?s Extreme PUE

ariel sabiguero yawelak asabigue at fing.edu.uy
Fri Apr 3 10:10:28 PDT 2009


I was wondering if the battery supplies energy for the whole system or
just the mainboard. If you check the cables, the hard-drives are still
12V+5V powered. Do you think that they just run without swap and if
power fails they just announce that nodes is going down and dies? Do you
prefer thinking that the power source has even more modifications and it
continues supplying 12+5V to the drives and keeps the whole system
running for a few minutes?
I don't where to put my money......

regards

ariel

David Mathog escribió:
> Nifty Tom Mitchell <niftyompi at niftyegg.com> wrote
>   
>> I suspect that the UPS component is very short.  i.e. finish the current
>> search, notify the distributed google world that the 'node' is going
>> off line and then go off line.  I doubt that the specifications for the
>> UPS are hours or days the way a national weather center or hospital data
>> center or other critical resource is specified.   If someone post a pool
>> for the UPS time factor put me down for 15min.
>>     
>
> Pretty much what I was thinking too.
>
> In the picture on the OP's link the external PS does not look any bigger
> than a standard unit, which leaves very little room in it for the UPS
> battery.  The primary purpose of the UPS is most likely just to give
> them a few minutes leeway to turn on the data center's generators when
> the main power fails.  Even a couple of minutes UPS coverage is plenty
> to allow them to save significant money on the backup
> generators/failover circuitry.  Also, a smaller battery wouldn't have
> the weight penalties of the sorts of sealed lead acid batteries usually
> found in a UPS.  And there could be some other advantages, for instance,
> they could throw the main breaker on a rack (assuming it is wired that
> way), move the common plug from one circuit to another, and then close
> the breaker again, all without losing service.  Big central UPS units
> don't provide that sort of flexibility.
>
> Regards,
>
> David Mathog
> mathog at caltech.edu
> Manager, Sequence Analysis Facility, Biology Division, Caltech
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
>
>   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.scyld.com/pipermail/beowulf/attachments/20090403/869648a7/attachment.html


More information about the Beowulf mailing list