"They all suck!" Re: UPS failure modes (was: fire at NAC)

Alex Rubenstein alex at nac.net
Fri May 30 00:29:29 UTC 2003



> We have two MGE 150KVA UPSes at our Newton facility. When I designed the
> electrical system, I originally specified a make-before-break 208V 450A 3PH
> switch.

What is the saying, the lawyer who represents himself has a fool for a
client??  :)



> This would enable us to isolate our internal and customer load
> panels from the two UPSes and their respective static transfer switches and
> output transformers - so we would have completely redundant dynamic A/B
> switching. The ONLY vendor at any price I could find was for a switch made

[... deletia ...]

After our investigations, our conclusion that paralleling was a costly
(pay for two UPS's, and essentially get one) and somewhat risky (it's been
demonstrated that paralleling makes a failure worse rather than better at
least some of the time) proposition. Plus, the added complexity of
installation (phase rotation?) and operation (phase synch?) is a little
too much for me.

Instead, we choose to have totally seperate power systems. For 90% of the
customers who don't care, and want a cheaper collocation cost, they are
happy knowing they live on one UPS.

For the guys who give a hoot, will pay money, and want it, we can bring
them another circuit, from another UPS. There is no inter-reliance on the
UPSs, and I feel more confident in the power I provide to the client.

I think that it is clear, if you really want the most reliability you can
ever obtain with sanity intact, and money is relatively not an object,
you'd have two grid feeds to the building; two ATS's, one per grid feed;
one generator per ATS; one UPS per ATS. Then, give each customer a power
feed from both grid feeds.

But no one, not you, not me, can do this and provide a $650/month rack
with 20 amps of power, and stay in business.


-- Alex Rubenstein, AR97, K2AHR, alex at nac.net, latency, Al Reuben --
--    Net Access Corporation, 800-NET-ME-36, http://www.nac.net   --





More information about the NANOG mailing list