Microsoft gets data centres powered up for big UPS turn-off The problem with data centres is that you have to plug them in. Or to put it another way, the problem with data centres is that they draw their power from a national grid and this ultimately narrows the cost efficiency structure under which they can operate. Conspicuously creaky and ageing grid networks in (even) advanced …
Google have been doing this for a few years: www.cnet.com/uk/news/google-uncloaks-once-secret-server-10209580/
"But the problem with fuel cells is they have to be run through Uninterruptible Power Supply units (UPSs) – and these cost big money and take up a lot of space."
Utter rubbish: http://www.bloomenergy.com/fuel-cell/mission-critical-data-center/
"A data centre powered by Bloom Energy eliminates the need for traditional backup equipment like diesel generators, UPS, batteries, and complex switchgear"
I mean, I thought you saved money by centralizing everything for ease of maintenance and control. Now it turns out you save money by distributing and avoiding infrastructure while benefiting from "commoditization".
It almost seems like whatever you are doing currently is wrong, and you should switch to the opposite while spending cash on some vendors new kit.
Marketing eh.
It's the same whatever you do, on average.
Centralising, decentralising.
Nationalising, privatising.
Thin client, fat client.
The cycle is endless as you can always find a benefit in one that doesn't exist in the other and so, over time, your prime perceived weakness appears to be addressed by the alternative - and it is. Until you then see that what wasn't a weakness before now is a weakness and, hey, that other alternative doesn't have that weakness... repeat ad infinitum.
Central UPS means greater efficiency and everyone gets the same crack of the whip but is large, specialised and can be a single point of failure. Distributed UPS is less efficient but cheaper with a higher maintenance cost as you run around random machines whose batteries die at random times. And so on.
In-computer UPS isn't new. Google have been doing it for years (but not on all of their machines). You can buy them for desktop PC's that go into the drive slots. Laptops effectively ARE their own UPS (and I've often used one just like that, powering down the other hardware knowing I have 2-3 hours of good charge before I need worry about the machine I'm doing it from).
All that changes here is a small standards-compliant version.
Let's take it to its logical conclusion. Why centralise all your data in the cloud? It makes far more sense to have your systems running close to where they're actually used. What's more, the more people who do it, the less the centralised data centres are needed, so everyone's a winner.
And they said I'd never be a marketeer...
for MS and google I'm sure its fine(for subsets of their workloads anyway), but for most folks these in server batteries don't provide enough run time to flip to generator. I want at least 10 minutes of run time at full load(with N+1 power), in the event automatic transfer to generator fails and someone has to flip the switch manually (same reason I won't put my equipment in facilities protected by flywheel UPS).
I heard a presentation on this kind of tech a few years ago and one of the key takeaways was that 90% of electrical disruptions last less than two seconds. Not willing to risk the other 10% myself.
I want at least 10 minutes of run time at full load(with N+1 power)
So do everyone, but perhaps it could be done with a PSU fitted with batteries...
If the server is powered by -48VDC (or 380VDC?) I believe the power conversion to DC would require significantly less space and perhaps one or two DC battery cells could be squeezed into the PSU.
Since the Server power management logic is aware of power outages it could start throttling the CPUs and other m/b electronics, slow down fans, allowing higher temperature for the few minutes the server will have until the batteries are depleted. Naturally all the PSU slots should be filled.
The downside would be the need to replace the PSUs every few years, or at least have the batteries replaced.
I know next to nothing about electricity so feel free to shoot the idea down! Sounds a wee bit too fantastic and I've probably missed something, like the laws of thermodynamics or something... :-)
The end conversion is more efficient but you're still pulling DC voltage from somewhere at some point - so you have to weigh up - efficiency when you're running on battery versus day-to-day efficiency. AC is, I believe, more efficient to transport around a building.
However, some datacenters and telephone exchanges have been on (as you point out) -48VDC for years. The hardware is available. Nobody uses it because it's not that much better than just a cheap PSU on a 240v line. And any electrician can come in to sort out the 240v, not all of them will touch deadly voltages at DC (hint: the reason we use AC is that, if it contacts you through Earth, you have a good chances of being able to let go as your muscles spasms rather than contract tightly as they do under DC - My father worked in garages that service lorries for decades, it's common practice to have a large piece of wood "on standby" should someone bridge the (DC) lorry batteries, because you literally have to whack their body away from it).
The fact that what you suggest doesn't happen en-masse and automatically everywhere tells you that there are problems with it. The circuitry to generate 240v AC from 12v DC is actually commodity hardware. The circuitry to distribute and step-down huge DC voltages is not.
And PSU UPS aren't unusual. Hell, you can get ones that go in empty drive bays and all sorts (don't Argos sell ones in the extension lead itself?). The "simplest" solution is often the best, even if it's the most technically complex. And 240v AC works everywhere and is available everywhere and doesn't require specialist rack units and "just works" and UPS to generate it are bog-standard.
AC is, I believe, more efficient to transport around a building.
It isn't. The inefficiency comes from power drop in the cables - which is determined by their resistance (effectively fixed) and the current flowing. It matters not whether it's AC or DC. The main reason AC is pumped around the building is that it has historically been very easy to change the AC voltage with minimal loss, whereas changing DC voltages has, until recently, involved much cost and power loss. It's probably still cheaper to change AC voltage, but the discrepancy is much reduced.
And any electrician can come in to sort out the 240v, not all of them will touch deadly voltages at DC
An electrician will not be worried about 50VDC. There's no more risk to life - in fact, probably less - than 240VAC.
The circuitry to generate 240v AC from 12v DC is actually commodity hardware. The circuitry to distribute and step-down huge DC voltages is not.
It's pretty much the same hardware. There's a small difference in the configuration. Maxim[1], for example, make devices that work in either direction within limits.
Vic.
[1] A friend of mine who's rather good with hardware design hates Maxim converter products because they don't vary their LO frequencies properly - they tend to pulse-skip. This leads to all sorts of nasty sub-harmonics that are more work to filter out. But I like Maxim because they give me free samples when I'm putting a new design together :-)
This post has been deleted by its author
As far as I recall lithium ion batteries really don't like deep discharge and wear out much faster than the equivalent sealed lead-acid.
So this technique means you'll be replacing more, smaller batteries more often - hoping that 1000 small batteries costs less than 10 big ones.
Not to mention the interesting fire loading - lithium battery fires cannot be doused, and the batteries catch fire quite easily when damaged or overcharged.
You basically have to wait for it to burn out and then put out the secondary fires.
Can their fire suppression system cope with that?
Actually if handled properly they can outlive a lead acid battery. You also forget the power density is much better with lithium and they can handle the load to a much lower state of charge than lead acid too. If you charge them to 80% they will last 4x longer than if you charge them to 100%. They also don't care what state of charge they're kept at where as lead acid must be kept at 100% to prevent sulfation of the cells.
It is true though that deep discharging any lead acid or lithium battery is bad for it, but lithium batteries nearly always have protection against this built into them as they become unstable if deep discharged. Similarly with over charging, they are often protected.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
