Slashdot Mirror
DC Power Saves 15% Energy and Cost @ Data Center
Krishna Dagli writes "Engineers at the Lawrence Berkeley National Laboratory and about 20 technology vendors this month will wrap up a demonstration that they said shows DC power distribution in the data center can save up to 15 percent or more on energy consumption and cost. The proof-of-concept program, set up at Sun Microsystems' Newark, Calif., facility, offered a side-by-side comparison of a traditional AC power system and a 380-volt DC distribution system, running on both Intel-based servers and Sun systems."
I, for one, would not be comfortable working around high power DC. Call me paranoid, but I rather enjoy my heart beating with its current interval. You can take all the precautions you want, but accidents do happen.
[sig]you really dont want the answers, trust me[/sig]
Read Stephen King's "Tommyknockers". You can do a lot of things if you go DC-only!
Where were you when the voynix came?
When they used to talk about DC power systems be less efficient, you have to remember that most of them were talking about 12-48 volt systems.
From the article:Not according to my electronics class, if they're really going to be running at 380 volts. They'll need more insulation instead. I'd also want to be real careful around those wires. DC will kill you much quicker than AC of the same voltage/amperage. Then again, you don't have to worry about shorting yourself to ground with DC.
For now.
I don't read AC A human right
15% seems compeling for DC power in new construction, but obviously this begs the question of switching costs. But 15% was just for the electricity used to power the servers, the article assumes as would I that there would be additional savings due to reduced cooling needs... that extra 15% electricity would have generated about that much heat. I'd like to see a breakdown of switching costs.
Telephone Companies had known this for years. This is why you can get 48vDC versions of most systems.
In a telephon e exchange 48v DC is the norm.
They have huge batteries and standy generators to keep the phone syste, running.
I'd rather be riding my '63 Triumph T120.
Good to see some more DC in use. Tesla was right about AC for many applications but DC has its merits and any useful application of DC is a credit to Edison's scientific achievements.
... those claims of saving "up to 15 percent or more".
That pretty much covers the entire range of possibilities.
I often wonder why they didn't say something like "up to 50 percent or more" or "up to 99 percent or more". Those would be every bit as meaningful.
Those who do study history are doomed to stand helplessly by while everyone else repeats it.
I always thought the opposite was true. Here is a wiki quote that also supports that:
Taken from http://en.wikipedia.org/wiki/War_of_Currents/
You can also store DC whereas you cannot store AC, meaning UPS always need an AC-DC followed by a DC-AC stage. Since we have had large FET power transistors it has been possible to make DC/DC conversion very efficient - especially since, if you were beginning again, you would not choose 50 or 60 Hz for best efficiency.
In fact, already the PC is using a DC bus to power small peripherals (USB) and it works surprisingly well.
I may be wrong about this, but it was Edison who accused DC power of being more dangerous ("Westinghoused") only to have AC adopted for the pleasant US custom of humanely frying criminals.
Pining for the fjords
Would be interesting to know what the efficiency is of a 380 -> 12/5 DC-DC converter, compared to a traditional 110 AC -> 12/5 DC converter. This is of course only just a part of the total picture, but in the past this has often been mentioned as the reason for _not_ going DC. Maybe with modern switching power supplies, that problem has disappeared.
Browsers shouldn't have a back button!! It's all about going forward...
This issue has a been a hot topic at conferences for data center professionals, with a lot of debate about timetables. Several facility designers are advocating DC distribution as the solution to the current power/cooling challenges. Corporate data center managers like the cost savings projections, but want to see it work in someone else's facility before they put their neck on the line and pitch a DC conversion to their bosses. That's the real value of the Livermore project discussed in TFA - it provides a working model.
Right now the cost of power is remaking the landscape of the data center industry. Yesterday there was another announcement of a huge data center in central Washington State. Sabey will invest $100 million in a facility right up the street from where Microsoft and Yahoo have data centers under construction. It's all about cheap hydro power. Both Microsoft and Yahoo have contracted for more than 40 megawatts of power from the local utility. That's why DC is one of the solutions that will begin to get serious consideration.
RichM
Data Center Knowledge
Agreed-- that level of DC voltage is actively dangerous, and I wonder why they are running that high a voltage when most of the equipment in the racks is gonna want +5VDC & +12VDC rails. Perhaps they are using it to power big DC motors in fans and AC/cooling/dehumidifying equipment, but it would seem to make more sense to feed those with 3-phase AC and use a more sensible VDC delivery at, say, 48V, which is a telco standard...
"The human race's favorite method for being in control of the facts is to ignore them." -Celia Green
Depending on where you are in the world 3 phase AC is 415V or 480V, and in industry we have no problem handling that. 380 VDC doesn't seem much of an issue to me with regards to insulation safety etc and I have dealt with control panels that have operator controls running at 240VDC (and grabbed them accidently and lived to tell the story) Though now days operator controls are being specced as 24VDC.
.. Bzzt .. Nope. grounding yourself is always an issue with ground referenced power systems. And I would never rely on any power system being perfectly isolated from ground. That sort of misguided thinking leads to nasty surprises.
But as for DC killing you quicker, I would disagree that its the type of system that kills you, it will depend on the type of damage that the shock causes. You can use a 9VDC battery to kill yourself if you apply it in a manner that a small current (mA level) flows to your heart and I would guess that the same level of AC current would also do the trick. On the other hand if you pass a large current through your body that causes physical damage (major burns etc) then it won't matter if its AC or DC if the so much of the body is destroyed as you will die eventually.
As for not worrying about grounding yourself with DC
I am Slashdot. Are you Slashdot as well?
While it has been rumored that Mr. Nikola Tesla is spinning in his grave
At 60 revolutions per second.
From the article:
(emphasis mine)
The power lost in the cables varies as the resistance of the cable and the current in the cable.
The power delivered to the equipment varies as the current in the cable and the voltage on the cable.
A 380 volt DC system can deliver as much power per unit current in the cables as a 380 volt AC system (assuming a near-unity power factor).
Ergo, the size of the cables for a 380VDC system will be the same as the size of cables for a 380VAC system.
So, if the comparison is against a 240VAC system, then a 380VDC system will have SMALLER cables, not larger. Only if the system being compared against is a 440VAC system will the cables be larger.
Also - a 380VRMS AC system will have a peak voltage of about 540 volts (two significant digits in, two significant digits out), and thus will require MORE insulation than a 380VDC system.
Also - the first things a switching power supply does is rectify the AC into DC and dump it into a capacitor (and usually do power factor correction): so a power supply designed to run from DC needs neither the power factor correction nor the big capacitor (a smaller cap will still be needed, but not one that can carry the system through the bulk of the AC cycle when the voltage is below peak). This makes the power supply simpler, and removes switching losses from the rectifier (granted, a modern synchronous rectifier based on IGBJTs will have a very low loss - but it still is a loss.)
Also - creating a backup for 380VDC is pretty easy - you use a battery bank floated at the 380VDC level. No need to "switch" from mains power to battery - you are ALWAYS running on battery, and the mains power is just charging the battery. This is how the phone company does it - the central office has a bank of batteries providing 48VDC, which is float charged from the mains. Lose mains power, and the system doesn't even blink.
(Yes, you need to have fusing to prevent those batteries from going nuclear if shorted, but that is a much simpler problem to solve than the issues of switching to backup power for an AC system.).
Yes, you have to design the equipment to run off the 380VDC - so you need different power supply front ends: most power supplies are split into 2 parts - the front end that takes mains power and makes about 300VDC on a cap, and the back end that makes the lower voltages from that - so the back end of the power supply does not need to be redesigned. Moreover, most power supplies use an off-the-shelf front end module, and any "magic" is in the back end - so this is NOT a major issue.
www.eFax.com are spammers
They use 380 volts instead of 48, because they can send almost 8 times the power through the same size wire.
I wonder if you would see the same 15% power saving if a home was outfitted for DC use ? When you think about it most electronic devices in the home have power supplies embedded which are nothing more than AC-DC coverters, which in and of themselves waste energy.
A DC power home would lend itself more readily to home based power generation. I believe most solar panels and windmill generate DC power which then has to be converted back to AC before it can be put on a powerline or used with conventional home appliances. With the new high efficiecy LED DC lights available the AC light bulb (a hundred year old device) is a real power hog and also generates enomrmous amounts of heat.
100 yrs ago when they were first bring electrical power to the masses perhaps AC was the right answer, but I believe our needs and priorities have changed in the past 100 yrs and perhaps the way we generate, distribute and use electricity is due for a new analysis.
1. DC/DC conversion is cheaper and simpler bacause with a 60Hz AC signal, you have *no* power during the zero crossing. The PS has to store the energy in a capacitor or a coil to deliver during the 120 "outages" a second. A DC/DC converter operates at hundreds of kHz, so components are much smaller, and since the conversion uses square waves, it does not have the "outages" a sine function has on the input.
2. A lot of AC/DC switching power supplies is a constant power load on the grid. It tends to draw more Amps as the Voltage decreases, producing a lot of harmonics in the mains power line, and a worse power factor than regular "resistive" equipment. Therefore the mains must be overdesigned to support this kind of load.
2. 220V AC means 220V *RMS*; 110V is just one of the wires tied to ground. The peak-peak is around 311V. Not that different from 380V
"Fix it"
Comment removed based on user account deletion
If a job's not worth doing, it's not worth doing right.
Copper losses are created by current and are described by the equation I^2 * R. So as you double your current, you quadruple your power losses.
Conversely, if you halve your current by boosting the voltage, you can reduce your transmission losses by 75%. Thats a pretty good reason to go with higher voltage. And since this is in the datacenter, you can train your people not to pee on the red wire.
Remember, You are unique...just like everyone else.
From the Merck Medical Manual "...The effects of AC on the body depend largely on the frequency. Low-frequency currents of 50 to 60 Hz (cycles/sec), which are commonly used, are usually more dangerous than high-frequency currents and are 3 to 5 times more dangerous than DC of the same voltage and amperage. DC tends to cause a convulsive contraction, often forcing the victim away from the current's source. AC at 60 Hz (household current) produces muscle tetany, often freezing the hand to the current's source; prolonged exposure may result, with severe burns if the voltage is high...."
Speaking of conductor sizes, the article said this:
A DC system also would mean having to bring in larger cables than now exist with AC power.
I challenge this notion. Conductor size is not related to whether the power is AC or DC or what frequency of AC it might be; it is related to current.
Larger cables are needed when more current is passed. Traditionally, you need larger cables for DC, because traditionally, DC power systems were lower voltages (12, 24, 48) than AC systems, and these lower voltages required larger currents for same power (e.g. 100W= 830mA at 120V, but 8.3A at 12V). Running at 380V, however, you get to lower the current (excluding the reduced current caused by the 15% power savings) versus a 120V system.
Expanding on that, the reduced conductor size is proportional to the square of the reduced current. Simply by going from 120V to 380V (a factor of 3.17), you change the current flow downward by a factor of .32. This means you can change the cable cross-section area to by a factor of .1; you reduce the cable to one-tenth its original size; one tenth the copper.
www.wavefront-av.com
I see their results, but I come to a different conclusion.
My headline would read "DC Power Results in 15% Increase in Equipment in Data Centers"
Don't hydrogen fuelcells generate DC? I can't wait to hook my home gaspipe to more than just my stove, and suck in the MWs direct to my devices. Without all those AC adapters left over from the 20th Century cluttering my home, getting hot, drawing power when "off", getting lost and mixed up...
Maybe we can use the old AC network as a 3rd broadband line, after telco and cableco.
--
make install -not war
They can deliver a DC-powered rack that will do the A/C conversion in the rack using a rectifier, so you save power by making the conversion only once. They also can take DC to the rack, and put pretty much whatever you want into the systems. Not to mention the high density you can get.
Blade enclosures also use a similar trick, the blades all get DC. And many data centers also have DC available already, you just have to ask.
http://www.rackable.com/
Instead of using a 380-volt DC distribution system, why not use the telephone standard: 48V? Then you could use the same 48V connectors (STD equipment, lower cost) to plug into converters that would be in the same form factor as regular AC power supplies. Then, if you really what to save money, have solar panels (or other means of power generation) to charge batteries that would feed the 48V system. Built-in UPS!
Hum, I might be too idealist, here...
Obama's legacy: (N)othing (S)ecure (A)nywhere and (T)error (S)imulation (A)dministration
Depends on the load. 48V isn't just used as the ring voltage on analog POTS lines, it's also commonly used as a power-delivery bus to PBX switches, SmartJacks and other CSU/DSU equipment for T1/T3/E1/etc lines, perhaps with a wall-mounted 48VDC battery backup unit.
Although, you're right that they don't use 22-gauge wire for that purpose; one of the PBXes at a client site has a 15 or 20amp/48VDC power supply, for example, which seemed to be using 14 gauge wiring, for example.
"The human race's favorite method for being in control of the facts is to ignore them." -Celia Green
- But they still need to have the transformers to step down the voltage
This is done with a pulse-width modulator. An AC-DC power supply already has one of these running from 380VDC anyway. The 380VDC in that case is derived from a type of rectifier called a voltage doubler (in the case of 120V sources) or a full-wave rectifier (in the case of 240V sources). The excess voltage then comes from the fact that we are getting peak, rather than RMS, voltage from the AC to the DC side.
The savings is in that the rectifiers are all consolidated. The pulse-width modulators can have an efficiency as high as 95% easily, whereas a whole switching PS can be as bad as 50% efficient.
The savings are in the economies of scale for the rectifier. A similar savings could be realised in the pulse-width modulator, too, but would be quickly wiped out by the increase in losses by making long wire runs at low voltages (5V and 12V).
- DC requires twice as many wires
Nope. Still two to complete a circuit, just like AC.
www.wavefront-av.com
To get AC, you spin a coil in a magnetic field.
To get DC you, um, spin a coil in a magnetic field, then rectify it, then put a huge capacitor on there to flatten out the humps.
There's just no good method for generating DC. And even if there were, electric companies aren't going to run two new phases (DC+ and DC-) to get it to you from the source.
Instead, the power is going to come to near you as 3-phase, then be rectified. There is a loss in that rectification, but sadly, you can't eliminate it, just change where it happens. Moving it to the other side of your power meter will have an advantage since you theoretically wouldn't have to pay for the losses, although the electric companies would surely change their rates to recoup this lost money. But note that even if they don't change their rates, you haven't saved any energy, just not paid for as much.
So my guess is this experiment bought into this fallacy, that they measured their power usage at DC levels, found it was lower and reported that as a win, when without a source of DC power that doesn't involve rectification it really isn't.
I'm sure they save some electricity due to the increased voltage. That reduces current, which decreases power lost. This is the same reason electric companies use high voltages for power transmission.
The article seems to imply that power supplies convert 120VAC to 381VDC internally. This just isn't true. They never raise the voltage, and 120VAC peaks at 175V or something like that. Even 240V input would peak at 350V. So I don't get this. I think they just messed up a few numbers and really in the experiment connected rectified 240V (UK 240V, which is one phase double high, not the US one 120V phase over another) directly into the power supplies after the point where the rectifier would normally be.
From what I can tell, going to DC just would save you the cost of lots of little rectifiers in favor of the cost of one big one. To be honest, since the small rectifiers come in commodity ATX power supplies, you're paying almost nothing for them anyway. So I don't see that it's all that valueable to consolidate them.
I would recommend that if we wanted to save the most power on servers, we should just go to 3-phase 440V AC power supplies. A new connector would have to be designed, as the current 440V 3-phase connector would barely fit on the back of a tower, and wouldn't fit on a 1U server. This would save the most possible in losses without having to buy external rectifiers or force the electric companies to install one on site (and charge you back in increased rates).
http://lkml.org/lkml/2005/8/20/95
"Soylent Green is IT people!"
So it's O.K. to pee on the black wire?
..........FULL STOP.
"And since this is in the datacenter, you can train your people not to pee on the red wire"
We have dedicated and colocated server in various datacenters, so I have a fair amount of experience with them, and so I need to ask you... PLEASE give me an example! An example of a datacenter staffed with people who can be trained not to pee on a red wire, because if they can be trained to do that... hell they might even be able to reboot the right machine from time to time!
The revolution will not be televised... but it will have a page on Wikipedia
That's true for transmission over longer distances, but what about those short distances in the data room? Or for that matter, in my home office?
Almost every device I own uses 4.7V or 12V. I look around at work here, and I can see power strips full of transformers, all of which are knocking back the AC power to one of a couple of DC voltage levels. Every one of those transformers has its own losses, most of which dissipates as heat. They're also large, making it difficult to fit them all into a strip, and their heavy, making it difficult to balance or hang the strip where it's most needed. At home I have DC transformers for the monitor, the switch, the firewall/router, the WiFi, the PDA recharger, the BT mouse recharging dock, the USB hub, the TV tuner box, etc, etc. It's got to be a safety hazard.
Then we have the PCs, which are also using 2 (OK, 4) predefined levels of DC voltage, and have their own transformers and rectifiers to do it. These get so hot they even need their own fans!
Why isn't this stuff standardized, and power strips can instead contain one single transformer/recitifer package, with DC sockets, or retractable DC wires coming out of them? Even if we ignored PCs and only did the external peripherals for now, we'd still get a big saving in power just by having fewer transformers.
That 48vdc comes from the central office, where thousands of amps of it are used to power the switch, all the transport gear, and most of the auxilliary equipment. (Air conditioning is all AC powered, but everything else runs from the central DC plant.)
The power conductors in central offices are oversized out of paranoia, and because sometimes you have a foot-thick pileup of power cables leaving a fuse bay and you want to make really sure resistive heating is negligible. Also, most equipment has redundant power feeds, A and B, but either feed is large enough to handle the entire load. During normal operation when both sides are sharing the load, the resistive drop in the wires is absurdly low.
The other advantage of 48v is that it's below the 50v "low voltage" standard in the NEC, which means it's easier, legally, to work with. The 300-plus voltage they're using in this study loses that advantage.
Also consider this: AC voltage and power are measured RMS, but the insulation has to withstand P-P voltage. So to deliever the same power on the same conductors, the DC system's insulation has a greater margin of safety.
The cynic in me suggests it's because your typical wall-wart costs about 50 cents to make in bulk and are commonly marked up by a factor of 20 to 100 or so, so when the company sells you a replacement they make out like Enron.
But yeah, standards exist-- most of the time, you can buy a generic PS from Radio Shack which delivers 3V, 5V, 7.5V, 9V, & 12V @ 1amp or so for much less than you can buy the product-specific wall-wart. Some vendors (like Sony) have even deliberately disregarded the JEDEC? standard connector sizes in order to prevent you from using a generic replacement PS.
"The human race's favorite method for being in control of the facts is to ignore them." -Celia Green
This seems to be the debate with Tesla, with the Intel vs Sun networking issues?
The freq switch smoothed out the motor in the tape drive. There was almost no hum detectable compared to 60 hertz. When listening to Soviet boats we would alter 60 hertz up or down (58, 59, 61, 62) to alter our own background noise and unmask tones on that line from the contact.
When operating on exercises with our own or allied forces we used a special transducer and amp to play tapes of various Soviet boats. I wish I had the 500watt McIntosh tube amp we used.
Professional Politicians are not the solution, they ARE the problem.
Stick a voltmeter across tip and ring. If the line is on-hook, you'll see 48V DC. If someone calls, 90V AC will be superimposed on it to run the ringer. Take the line off-hook and the voltage goes down to somewhere around 6V DC.
20 January 2017: the End of an Error.
No, he just had good marketing. People keep repeating a lot of things he never actually invented, just proposed some vague idea, together with other things that proved totally impractical. I wonder why they never quote this:
"The aeroplane is fatally defective. It is merely a toy-a sporting play-thing. It can never become commercially practical. It has fatal defects."
Nikola Tesla
Edison, it's true, had his personality defects. In many cases he was not quite ethical. He was wrong in trying to push DC technology at that time, although with modern electronic components transmission of DC at very high voltages is not only possible but the only practical alternative in many cases.
Now consider Tesla's idea of bladeless turbines for instance, impractical at the time, still impractical today. In order to work with reasonable efficiency, it needed a set of disks with a tenth of a millimiter in thickness separated by less than a half millimeter. Try sending a flow of high pressure steam into such a machine, a fraction of a second later you'll have a big mess of crumpled metal foil. Or how about what is possibly Tesla's most famous experiment, a wireless system for transmitting electric power. It radiated energy away in all directions, to be received here and there by special antennas. With luck this system could use maybe a fraction of a millionth of the total power transmitted. A few cities in the early 20th century had Edison's impractical DC power distribution system, but Tesla's wireless system was never adopted anywhere because it was so extremely inefficient.
The main difference between Edison and Tesla is that Edison made research with practical applications in mind. Tesla proposed many ideas without even building prototypes, Edison made painstaking trials until he got it working. For instance, in his incandescent light bulb, Edison's lab tested more than a thousand different filament types before finding one that worked. Edison's main invention that Tesla ignored was a system for developing applications for new ideas. It's not enough to have brilliant ideas like Tesla did, one must put those ideas to practice, like Edison did.
Forget DataCenters.
How about 12V DC at home, so that I can get rid of the half dozen little tranformers plugged-in under my desk?
What are the real objections to 12V in the home?
The most sensible one I've heard is most digital devices need a fairly fixed voltage drop, which probably means that need to be in parallel. That pushs your current thru the entire system up, perhaps to dangerous levels.
Have peed on a few electric fences. One I felt. I think there are conditions where when the right things line up, you get a numb pecker.
Politics is the art of looking for trouble, finding it everywhere, diagnosing it incorrectly and applying the wrong fix.
What I could see happenning, though, is special small size switching transformers built into a standard electrical junction box, which are "smart" in some manner to know when a plug is plugged into them, which connects the switching transformer in, and that supplies, say, 12V at 10A or something to a common downconvertor system or something that all the other peripherals plug into (that, or each peripheral converts the 12VDC independently). In a way, I built something like this, once, for a desk I had: I hooked up an old Sun Computer pizza-box (Sparcstation?) powersupply and created a "bus" of electrical wires running under the desk, hooked up to screw terminal bus strips every so often. I ran the 12V, 5V, and ground lines via this bus along the length of the desk, so then I could get 12V, 7V, and 5V feeds from this system. Hooked up all my peripherals that had wall warts to the bus, and ran a "power on" green LED to the front of the desk for status. Worked pretty well.
Reason is the Path to God - Anon
Telecom companies have been using DC distribution systems for DECADES because they don't have to lose energy converting back and forth between AC and DC. It's about time the computer industry is catching on.