For decades, Post Office Telephones > BT managed to maintain systems that were so good, they even worked when all the power was out to your street, yes, the telephone line even supplied DC power to make the handset work, and of course the exchanges themselves were solid DC power backed by deep cycle lead acid battery banks and big solid bus bars to carry the current with minimal voltage drop.
I’ve built a lot of custom “server” equipment for people over the years, none of it ever died or failed to work 24/7/365, even when components failed, it continued to work and those components got swapped out.
One of the reasons I always claimed that this was so was because I used very expensive (comparably) DC-DC PSU’s and fed them off a nominal 24 VDC rail, and generally speaking the DC supply was two 120 ampere hour deep cycle batteries in series per “pc/server box”.
The math is fairly simple, series batteries and the a/h rating stays the same but the voltage goes up, so at a ten hour rate 2 x 120 Ah batteries give you 24 VDC and 120 Ah, and since it’s all DC watts = volts x amps so you’re looking at 24 x 12 = 288 watts that you can draw for 10 hours, and 288 watts will run a dual CPU headless server box quite nicely thank you.
That wouldn’t have been enough for the old Post Office, they wanted 24 hours of battery life, which gave enough time to rebuild the DC gen sets to recharge the batteries, and once they were running the entire phone system could stay up indefinitely…. and they would have wanted 48 volts because it halves the current and decreased voltage drop and losses…. nor would they use cheap shit truck/boat batteries, they’d all be fuck off big slabs of 2 volt traction cells
Nor should anyone under estimate just how much juice a national phone system can draw, I was shown the national grid power feeds for the days the planes flew into the WTC buildings, and everyone picked up a phone… 10% of Uk power generation was going into the phone network.
Nor do you have to sync a DC gen set / charger to the load, it’s DC, so you just have banks of fuck off big heatsinked diodes, as soon as your charging circuit voltage peaks over bus voltage charging current will flow, it’s magic.
You don’t have a UPS, you have fuck off big banks of lead acid with capacity measured in tens or hundreds of thousands of ampere hours and it’s all glorious DC and it’s all you can do to push it hard enough to get a ripple out of it, and if you stayed 48 volts or less you *shouldn’t* get stuck to it, it could happen, but it was *less* likely, by the time you got to 110 VDC or higher banks it was 50/50 depending on which muscles locked up hard enough to tear ligaments or even break bones…
Then came the fucking AC UPS, the power came from lead acid basically, but the endurance / run time was pitiful, and this applied both to installed capacity as well as the electronics, 100% duty cycles measured in minutes at best.
The only ones that were *any* good were line interactive ones, where the output AC was *always* generated inside the UPS, and that power either came from a mains supply or the pitiful lead acid backup, and then you have synchronisation with the mains clock, again trivial with electronics.
So today in the Register, a FUCKING TECH JOURNAL, they are talking about the BA fuckup and saying the latest shit is that someone reconnected a UPS without switching out the mains or syncing first, and 240 VAC + 240 VAC = 480 volts so it’s no fucking wonder everything fried.
WHAT THE FUCKING FUCK????!!!
You’re supposed to be a fucking TECH journal.
240 VAC + out of phase 240 VAC won’t give any more than 240 VAC, what it will give you is YUUGE BIGLY current flows as the national grid AC clock and the device AC clock pull at each other, in rotating equipment it can literally rip things apart of off their mountings, i’ve seen it happen, with solid state equipment like a UPS you get to find what happens when 100x max rated current starts pushing through everything before breakers can trip, the magic smoke comes out, so does the more rarely seen magic light and magic sound.
But, even if all this shit is true, it still only explains ONE DC going down bigtime, it does not explain the second one, nor does it explain global ops going down.
I can tell you for a fact, you *can* still go out and spend US$20k on a UPS that is all controlled by buttons or a remote interface, and it is still possible to tell it to link into the mains clock without syncing first, it should be hard wired fucking impossible to do this… it would take at max less than 0.02 seconds for three separate “all must agree or none shall pass” circuits to sample the mains clock before spinning up the UPS power transistor control clock circuit, and bake it in at the circuit level so nobody can bypass it even if they tried.
But it still doesn’t explain the other DC or the global network.
Only ONE thing does, and it is the ONLY thing that is 100% backed up by the fact that to date no whistle blower has contacted the press with the inside skinny.
That thing is a cascade failure that happened because a whole raft of supposedly independent but nevertheless connected departments and procedures have all been operating at “fuck it, do NOT open the bonnet or touch anything, some how the fucker is still working” levels, and NOBODY is coming forwards because NONE of those various departments have clean hands.
But hey, it’s *completely* fucking separate to air traffic control or any other safety critical systems, no connection whatsoever, that is not just a different world, that is a different universe.
And if you buy that, I have some moon dust to sell you.