[Seattle-SAGE] Q: EPO (Emergency Power Cutoff Switches), data center design question.

Jeff Silverman jeff at commercialventvac.com
Wed May 25 17:14:37 PDT 2005


Last time I designed such a system, the power went through the contacts 
of a relay. When the relay was energized, then the current flowed. 
Pushing the big red button opened the circuit to the relay. The system 
was also wiring to the fire alarm so that if a fire was detected, the 
power would drop automatically.

A more sophisticated system might introduce a delay of, say, 5 seconds, 
to allow systems to shutdown gracefully, e.g. sync; halt. Pushing the 
big red button asserts the power fail signal, and then after the delay, 
de-energizes the relay.

The important design principle is that the system should be fail safe, 
i.e. any failure mode in the system will cause the system to fail in a 
safe way. About the only thing that can go wrong with the system as I 
described it is that the relay can freeze in the closed position. Other 
than that, if the wire breaks, if the coil of the relay burns out, if a 
fire burns up a wire, if there is a more global power failure, then the 
system will shutdown.

Note that you might want to have disconnects both "upstream" and 
"downstream" from the UPSes. If the power going to the UPS fails during 
a fire, then the UPS will continue to power the equipment and you have a 
"Class B" (Electrical) fire. However, if you disconnect the UPS as well, 
then the fire becomes a "Class A" fire, which is much easier and safer 
to extinguish.

In any event, the plans should be run past an electrical engineer or a 
fire engineer before implementation.

Incidentally, the Big Red Button is useful when all sorts of things go 
wrong. For example, I once had a 1" water pipe burst directly above a 
rack of USAF equipment. I hit the big red button and then moved the rack 
out of the way, getting drenched in the process. But because the power 
was off, I was safe. You can wire a thermostat in series with The Big 
Red Button and set it for, say, 95 degrees F and that will protect your 
equipment if the cooling fails. You can wire a moisture detector in 
series with the Big Red Button and that will detect flooding.

This was a terrific question.

I hope this helps.


Jeff Silverman

J. Nyhuis wrote:

>
> I am in the process of designing a small data center and am trying to 
> spec out a EPO Switch (The big red button that kills all power to the 
> racks in the case on an emergency).
>
> The old fashioned method was to use a huge breaker-like device that 
> physically cut the power to the racks. The vendor I recently contacted 
> is claiming that it's not done that way anymore. The EPO they want to 
> sell me triggers a circuit to each "EPO enabled device" telling the 
> device (such as a UPS or smart PDU) to shut down nicely, without ever 
> actually cutting the power to the racks.
>
> I can foresee several problems with this system, such as:
> If there is an emergency, (e.g. the PDUs are on fire), what's the 
> likelihood that the "shut down nicely" circuit will still be 
> operational and will work? (Usually there is a pull on the UPS to shut 
> down the UPS in an emergency fashion). According to the manual I 
> downloaded, the communication between the EPO and "smart EPO enabled 
> devices" is over cat5 or 24 gage wire, and I can tell you from 
> experience that cat5 does not survive long in fires... ^_^
>
> Has anyone designed a data center recently? Would you share how you 
> met this fire code requirement? Is this vendor selling me a line of b#ll?
>
> "***The EPO switch will not directly break the 208V circuits. There is
> too much power to run through the switch. It is designed to operate a
> contactor or other relatively low level device that in turn will control
> the actual power equipment."
>
> Sorry, as a sysadmin, I just see more single points of failure in an 
> emergency situation... ^_^
>
> Thanks in advance for sharing any advice,
>
> John H. Nyhuis
> Sr. Computer Specialist
> Dept. of Pediatrics
> HS RR349B, Box 356320
> University of Washington
> Desk: (206)-685-3884
> cabal at u.washington.edu
>
> _______________________________________________
> Members mailing list
> members at lists.seattle-sage.org
> http://lists.seattle-sage.org/mailman/listinfo/members





More information about the Members mailing list