NSA's New Utah Data Center Suffering Meltdowns 241
linuxwrangler writes "NSA's new Utah data-center has been suffering numerous power-surges that have caused as much as $100,000 damage per event. The root cause is 'not yet sufficiently understood' but is suspected to relate to the site's 'inability to simultaneously run computers and keep them cool.' Frustrating the analysis and repair are 'incomplete information about the design of the electrical system' and the fact that "regular quality controls in design and construction were bypassed in an effort to fast track the Utah project."" Ars Technica has a short article, too, as does ITworld.
Iranian Stuxnet? (Score:5, Interesting)
Hmm...so your equipment is randomly failing...you don't say?
Power management (Score:5, Interesting)
They probably used a power budget similar to the public Facebook datacenter data but then decided to run their machines on Windows Azure.
I have noticed that power consumption of my computers is significantly higher when running Windows - and the laptops have seriously reduced battery life, even while doing nothing.
Re:good? (Score:2, Interesting)
The NSA doesn't need you or I or our children to spy for them. Everything they want to know is willingly provided to them, by us, in real time. The Stasi had informants everywhere, which put people on their guard. Most people today don't think twice about saying things, because there are no daily reminders that somebody could be listening. That's far more frightening, in my book.
Switch-mode power supplies (Score:4, Interesting)
A friend of mine is researching power surges in the local town.
Most building codes under-specify the gauge of the neutral/return wire. For illustration, if you have three phases each rated for N amps, there is one shared neutral/return wire rated at N amps going out. At the end of the runs all phases are connected to the shared neutral line.
This is due to the nature of 3-phase electricity: the phases will tend to cancel out, so in a perfect setup you would need no neutral/return at all. Of course, the load on each phase won't exactly balance, and the load can vary as people connect/disconnect appliances, so you still need the neutral line in practice.
(Not true for house wiring, which has one or two phases coming in. Each phase has a return with the same gauge as the supply.)
This was fine when appliances were (generally) resistive loads, but nowadays switch-mode power supplies are common. When you do some math, it turns out that this type of load appears equivalent to 120 Hz power coming together at the neutral/return junction. Since 120 Hz [equivalent] power does not cancel out, the power in the return wire can be 3x as large as the building codes allow.
I've got a book explaining all this. Typically the neutral line will heat up and catch fire, breaking the circuit. Once that happens the various phases are connected without a neutral, playing hob with whatever is on those circuits and making occasional high-power ground loops and other unexpected behaviours.
Re:Iranian Stuxnet? (Score:4, Interesting)
65MW is a hell of a lot more than a small city, it's enough for ~65,000 average homes which is ~180k residents which would put it just outside the top 100 in the US.