The most common cause of downtime for a data center isn’t problems with power. It’s not overheating, and it’s certainly not environmental hazards. As is the case with many IT problems, the most common cause of data center downtime is, of course, human error. We have a way of taking something that’s humming right along and bringing it to a halt.
One of the most vulnerable points for human error in the data center is the EPO switch or “Emergency Power Off” switch. Understanding why the EPO switch exists and how it’s most commonly misused can help you avoid potential downtime.
Why do we have an EPO?
The obvious question is, if the EPO switch is a danger to data centers, why do we have them? The answer goes back quite a few years, and is a tale of regulatory control.
There was a time when mainframe computers who used tremendous amounts of power could find themselves in a burn-out situation. When that happened, it was better to have the mainframe crash than to let it burn out, mainly due to the potential loss of capital investment. The EPO shuts down the entire data center, giving a hard crash to everything inside.
The National Electric Code introduced the EPO switch in 1968. At that time, a standard was created where the EPO switch was necessary, and often at each exit door to a “information technology equipment room. This, then, creates several points of failure where human error could conceivably shut down the entire data center.
Protecting today’s data center from the EPO
It made sense, in 1968, to have an EPO switch. Today, however, the benefits aren’t nearly as clear. Yet, many local code authorities or building inspectors may demand that an EPO switch be clearly visible. And, while you could put the switch into an emergency “break glass” box, this isn’t compliant with the code.
What you can do is install a transparent box with a lift cover that also kicks off an alarm when the cover is raised. This should prevent most accidental activations of the EPO switch. At the very least, it should deter the curious passer-by from bringing the entire data center to a halt.