UPS - single points of failure

[mattboston at gmail.com (Matt Shields)]

If you want to eliminate that single point of failure purchase 2 UPS's
and buy a server has has 2 power supplies.  One goes into each UPS.

On 12/1/06, Rich Braun <richb at pioneer.ci.net> wrote:
> (This is a resubmission, the first one went into the BLU void.  My comments
> remain unchanged four days later.)
>
> Date:   Mon, November 27, 2006 10:54 am
> To:     discuss at blu.org
>
> As I prepared Thanksgiving dinner, the UPS on my file server abruptly gave up
> the ghost after a few years of service.  I'm obsessed with fault tolerance and
> power efficiency, so the "Uninterruptable" Power Supply has always been a bit
> of a sore topic for me.
>
> (This is a bit long, executive summary is:  consider Xantrex Powersource 400
> because it provides long runtime, 10 times longer than rivals.)
>
> In the past 10 years, I've seen the quality of hard drives increase to the
> point where they are more reliable than UPS units.  (When was the last time
> you had a hard drive die?)  Yet consumers still buy hard drives--a product of
> high complexity--with the expectation that they'll inevitably fail someday.
> Consumers buy UPS units--fairly simple products, by comparison--with the
> expectation that they should never fail, at least until the batteries give
> out.  For some reason, the marketplace has not led to the reliability
> improvements one would expect in UPS units.  A decade ago, I recall that many
> of the customer-visible outages at the ISPs I helped to manage were caused by
> failed UPS units.  Problems persist today.
>
> APC still has the bulk of the market for consumer UPS units.  I absolutely
> *hate* that brand, at least in the consumer category.  So when my Tripp-Lite
> failed, I sought anything but APC.
>
> I'm writing this because I stumbled into a unit that, if it proves reliable,
> is a surprisingly good deal versus the consumer-grade competition.  At my
> house, power outages are relatively uncommon.  There are maybe 10 brief (less
> than 10-second) interruptions annually, and maybe 1 or two lengthy (1- to
> 5-hour) annually.  Interruptions of 10 seconds to 3600 seconds almost *never*
> happen (maybe once or twice a decade), and I haven't seen a super-long one
> (6-hour plus).  Hence for me the typical runtime spec of a UPS unit is
> inappropriate to my needs.  A minute or two of runtime is enough avoid crashes
> during short interruptions and/or to save my work; 20 or 30 minutes of runtime
> isn't going to last through a typical long outage, and isn't worth paying
> extra for.
>
> There's a manufacturer of renewable-energy products which has come out with a
> UPS that seems to fit my needs just fine.  I got a couple of these units
> (Xantrex 400) at Best Buy this weekend, after comparing online prices and
> Microcenter (they have the brand but none in stock, at a price higher than
> Best Buy).  Last night I ran a runtime test.  Get this--my file server ran 6
> hours 55 minutes before the unit shut down!
>
> My VIA-based server with 2x300Gb storage only consumes 48 watts but this
> implies that even a 100-watt server could run for over 3 hours, thanks to the
> big honkin' batteries inside (480 volt-amp-hours' worth.) If you had to buy
> the batteries separately, you'd pay about $120--this UPS sells for not a whole
> lot more than the batteries alone.
>
> On the efficiency side, the annual operating cost of a UPS here in New England
> is about $1.75 times the wattage differential between input and output.  My
> old Tripp-Lite ate about 11 watts along the way from the wall to the server.
> The Xantrex uses about 5 watts.  So I'm also saving about $10/year with this
> unit installed.
>
> Why am I writing this review here for the BLU group as a suggestion for
> consumer (versus data center) installation?  Because in a data center you'd
> probably want to have the UPS units connected to monitoring software.  Even
> the cheap consumer grade UPS units usually have a USB or serial port.  Xantech
> omitted this essential ingredient.  Were I to design this thing, I'd leave out
> the silly LED panel display and include USB, which would probably reduce build
> cost anyway.
>
> Your mileage may vary, and I might ultimately be an unhappy camper if this
> product turns out to be an unreliable turkey.  But I'm very happy to see long
> runtime arriving in a consumer UPS product (given that these products, and PC
> power supplies, are promoted mainly with useless/misleading wattage figures
> designed to appeal to adolescents seeking "more power").  Might be worth
> contemplating for your own Linux setup.
>
> Now, someday maybe the power supply manufacturers will come up with a way to
> make consumer-priced PCs without a single-point-of-failure on the power input.
> I guess that'll be the day when Microsoft includes software RAID1 in their
> "Home Edition" O/S. ;-)
>
> -rich
>
>
> --
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean.
>
> _______________________________________________
> Discuss mailing list
> Discuss at blu.org
> http://lists.blu.org/mailman/listinfo/discuss
>

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.