Are Data Centers Finally Ready For DC Power?

1sockchuck writes "It's been five years since a landmark study outlined the potential benefits of DC power distribution in data centers. But adoption of DC in data centers remains limited, even as the industry aggressively pursues a wide array of other energy savings strategies. Advocates of DC distribution are hoping a new study will jump start the conversation about DC distribution, which can save energy by eliminating several wasteful AC-to-DC conversions within a data center. Meanwhile, an industry association for DC power adoption, the EMerge Alliance, has formed a new technical standards committee for data centers, and is advancing a 380-volt DC power standard. Will DC distribution ever gain momentum in data centers?"

Comments

Edison reaching out from beyond the grave

[ravenspear]

I told you bitches I would prevail one day!

Re:Edison reaching out from beyond the grave

[rolfwind]

Not really. AC was the answer to how to transport electricity long distances.

Currently, it is still converted to DC in a huge amount of devices, in the computer at the PSU. Few devices use AC iirc, something like a fan/ceiling fan probably has an AC motor because a DC motor would slice your finger off if you decided to play with the blades. So, the question then just remains how to optimize the point of conversion. It's rather like the electric car-fossil-fuel-electric-plant/gasoline car debate: have a bunch of small inefficient combustion engines or a large efficient one but deal with transport losses (along with a bunch of other issues).

In this case, just where along the line do you convert the AC to DC. Since DC can't really travel far at all without significant losses, I guess that would be at the rack level?

Re:Edison reaching out from beyond the grave

[Wonko the Sane]

AC was the answer to how to transport electricity long distances

AC was used because using transformers to convert between voltage levels was more efficient than motor-generators and solid state electronics hadn't been invented yet. All other things being equal, DC is always more efficient than AC for long distance transmission.

Re:Edison reaching out from beyond the grave

[Elder Entropist]

Very high voltage was the answer to how to transport electricity long distances. AC was the answer to how to convert that high voltage to safer/useful low voltages cheaply. Very high voltage DC can lose less power over distance than AC. On smaller/cheaper wires too due to the AC skin effect.

Re:Edison reaching out from beyond the grave

[Mr Z]

The problem with DC back in Edison's day was that you couldn't easily step it up or down. DC doesn't have higher losses than AC at the same voltage. In fact, DC radiates less energy away than AC does, and is therefore more efficient.

Ohmic losses all come down to I^2 * R. R is the resistance of the cable, and I is current. To deliver a given amount of power, you have to have a certain V*I. To reduce Ohmic losses, then, you have to reduce the amount of current, which means going up in voltage.

Incidentally, that's also what's driving automobile manufacturers toward 48v instead of 12v, since it would cut the current from the battery by a factor of 4, thereby reducing the amount of loss in the wiring by a factor of 16. That means you can use smaller wires to deliver the same amount of power, safely.

Re:Edison reaching out from beyond the grave

[khallow]

Since DC can't really travel far at all without significant losses, I guess that would be at the rack level?

Transmission losses are actually less than for AC. They don't lose energy to inductance with nearby conducting loops and impedance losses are about the same as for a three phase line with same RMS voltage. The real problems are conversion to and from AC, and the fact that DC operates at a much lower voltage (low voltage results in high losses, whether AC or DC) when in actual servers.

The idea behind DC powered centers is that the AC to DC conversion is done in one place, away from the servers so that a) the heating load of the center is lower, and b) it can be done in one place with a relatively efficient converter rather than in a thousand places with less efficient designs. The accompanying baggage as I gather is that you're either running a lot of power-losing low voltage lines or doing some sort of power-losing DC voltage step down inside the center.

Re:Edison reaching out from beyond the grave

[Anonymous Coward]

Way to take a completely normal, productive conversation and shit all over it.

Try being less of an asshole for a day or two. You might find that people hate you less.

Re:Edison reaching out from beyond the grave

[EETech1]

That's why they use 380 volts! One big splice goes to all the 12 volt stuff, then another splice comes off of that splice to do the 5 volt stuff. It is not run through regulators, it just happens automatically due to the superior characteristics of DC power! They also tap into the ground wire at various places to get the -5 and -12. Magic I tell ya!

It sees it's best efficiencies running near 100% utilization through so you want to plan your workloads accordingly, or you risk watching your $#!+ let out it's magic smoke! So all in all it should drive down the price of "the cloud" by forcing competition!

Win, Win!

Re:Edison reaching out from beyond the grave

[Relayman]

At 60 Hz, the skin effect is virtually nonexistent.

Re:Edison reaching out from beyond the grave

[Anne Thwacks]

The skin effect is only significant at frequencies WAY above 60Hz, so this is not relevant. AC is used because transformers can convert huge amounts of power from one voltage to another reliably, efficiently and cheaply.

Its true you can convert voltages DC to DC using electronics, but reliably at 132kV? It aint easy, cheap and certainly neither if you want efficiency.

(Blow up a 132kV IGBT, then come back and tell me about the damage :-)

Re:Edison reaching out from beyond the grave

[WaffleMonster]

I told you bitches I would prevail one day!

There seems to be a popular/fundemental misunderstanding of the tesla/edison debate.

DC is MORE effecient on the wire than AC given the same voltage, amperage and wire gauge.

The reason for this is in AC systems eddy currents induced by changing electric fields at 50/60hz cause electrons to migrate away from the core effectivly reducing wire size.

Why AC has been the choice for so long is an engineering problem.

Building rectifiers to convert AC to DC from huge AC generators which produce virtually all of our electricity with the kinds of voltages needed to carry massive quantities of volumes of energy is difficult, unreliable and ineffecient..even today.

Back then it was practically impossible. The choice between Tesla and Edison really boiled down to high vs low voltage. Low voltage transmission required impossible quantities of copper or decentralized generation.

Tesla wanted larger more centralized generation which given what we use for fuel these days is an exceedingly smart move.

Re:Edison reaching out from beyond the grave

[ssyladin]

As an aside, car manufacturers are also moving towards higher voltage because that gives them easier access to drive high-power systems like power steering. Currently most power steering is pneumatically driven, which comes with a hefty overhead cost in terms of manufacturing and maintenance. With a high-voltage bus to drive it, the complex machinery simplifies it to an electric motor and some gears.

Re:Edison reaching out from beyond the grave

[lgw]

That "whooshing" sound you hear? Yeah, the guy who modded that post "informative" heard it too.

Re:Edison reaching out from beyond the grave

[Anonymous Coward]

At 60 Hz, the skin effect is virtually nonexistent.

Monster Cable begs to differ!

Re:Edison reaching out from beyond the grave

[ktbaia]

I beg to differ. The Electronic Power Steering in my 1995 NSX decidedly does NOT feel like crap. When EPS systems were first put on production cars (the NSX was the first), it was on sports cars, where good steering feel is very important. See this link for information on how a decent EPS system should work. Just because Toyota can't figure out how to do a decent EPS doesn't mean that it hasn't been done. http://www.nsxprime.com/FAQ/Technical/eps.htm

Re:Edison reaching out from beyond the grave

[bkcallahan]

A/C skin effect applies *based on frequency*; I assure you skin effect at 50/60Hz, even at high voltages, is negligible -- Doesn't matter if it's 420kV or 4.2V. A circuit at 29.350 MHz at low voltages has to worry a LOT more about it than a 50/60Hz line voltage. (And it really starts kicking off at the start of the microwave range, 300MHz, and is exceptionally important by the end of the microwave range 3GHz.) The reason A/C is used is based off of Ohm's law and is based on current and resistance; Jack up the voltage 1,000x and you can reduce the current (and therefore, heat loss) 1,000x.

also needed for houses

[Anonymous Coward]

How many little wall-warts does the average house have? Tens? We need low voltage DC in our houses, and standardize all the little widgets on one of (say) two voltages. Each outlet could supply them in a dedicated connector alongside the current AC.

Re:also needed for houses

[LehiNephi]

I do/have done both (run electrical and understand the implications), and GP has a point. When I think of the things in our house that *must* run on AC, it's only our fridge, freezer, and HVAC. Everything else in the house either converts it to DC or could run quite happily on DC. For certain you'd want to have a different kind of plug for DC devices, but even that would give us an opportunity to 1) standardize on one global plug standard, at least for DC, and 2) allow us to design a small, rugged, safe type of plug.

Or is there some implication that I'm missing, and that you decided not to point out, in favor of flaming GP?

Re:also needed for houses

[vlm]

For certain you'd want to have a different kind of plug for DC devices, but even that would give us an opportunity to 1) standardize on one global plug standard, at least for DC, and 2) allow us to design a small, rugged, safe type of plug.

Aka the famous (in some circles) Anderson Power Pole. Go ask a ham radio guy.

The thing I love about in house DC distribution, which I have in my house, is it forces at least a token effect at "green power reduction". Suddenly given the choice of a 12 volt 6 watt LED fed by $2 of small gauge wire vs something resembling welding cable wire to run a 200 watt halogen, you make the ecologically correct choice.

I used to use cast off surplus 200 watt desktops for my mythtv frontends. Unholy pain to run on 12 V. Now I use 5 watt Zotac boxes. Good for everyone in every way.

Re:also needed for houses

[Obfuscant]

What bothers me is all the new LED bulbs that have transformers in them (guessing, because they get hot! ... feels like wasted energy)

High power LEDs get hot because you are running good amounts of current through them, not because there is a transformer. Transformers are pretty much useless with the DC current that runs LEDs.

I'd think it would be more efficient to run DC to lighting and certain outlets like those where small devices would sit ...

The problem comes in deciding what voltage to use. 12V means you need rather hefty wires to get the required current for some devices. A 6W LED needs half an amp at 12V. If you use a voltage that makes the current resonable, then you need to convert that voltage to what your device needs, every place you have a device.

Sending 380V means you can use the same or smaller wires than you'd use for 120V systems, but you'll be busy converting that 380V DC to 12V DC or 5V DC or 1.2V DC -- and while DC-DC conversion has gotten a lot better, it is still more complicated than a simple transformer.

Telecom's been doing this for many, many years.

[Anonymous Coward]

DC power is the standard in the telecom industry.

I design systems based around HP's BladeSystem, and the DC power modules just drop in and go. It's very easy, works great, and most of all, my telecom customers love them.

Re:Telecom's been doing this for many, many years.

[Shatrat]

I work with DC power in Telecom and it has 3 huge advantages I can think of off the top of my head:

1) You centralize your rectification. Instead of having hundreds of power supplies running at 80% efficiency, you can have a large rectifier system running at up to 96%.
2) Lead Acid batteries are hugely more reliable and less expensive than equivalent UPS systems, and provide more holdover time. They're still expensive and finicky, but many times less so than a UPS.
3) Any old technician with a brain in their head can run DC power feeds to equipment relatively safely due to the low voltages involved. AC power work of any kind should have a qualified electrician involved.

Re:Telecom's been doing this for many, many years.

[the linux geek]

Yeah, I've never quite figured out why telecoms have standardized on 48VDC while everyone else completely ignores its existence. Most midrange servers (HP Integrity and Nonstop, iirc most smaller SPARC Enterprise boxes, some commodity stuff) are available in 48VDC configurations, so it's not like there's a lack of hardware for it.

Re:Telecom's been doing this for many, many years.

[Anonymous Coward]

The positive ground in telco systems is not bizarre at all: one end of the twisted pair is grounded, and, being at zero volts does not suffer galvanic corrosion. The other end is at -48V and benefits from cathodic protection: it's the anode that generally gets corroded in a galvanic cell.

Re:Telecom's been doing this for many, many years.

[vlm]

4) Done right with a positive ground system, leads to less corrosion problems with outside plant. Admittedly "inside" the data center, if you're got corrosion, you're doin it wrong.

5) Less AC hum. We had some microwave site to site short hop gear back in ye olde NTSC days that could only be run off battery without 60 hz interference bars on the screen. Not technologically relevant anymore, but the point remains that DC is always going to be cleaner than AC.

6) Better lightning protection. I'm sure its happened, but I've never heard of losing a telco DC bus. Big conductors, giant batteries across them, lightning is just not an issue anymore at the power level (still need to ground feedlines / waveguide / whatever you've got at home like that)

7) dump most of the power conversion heat in the battery room where its all built to handle high temp and no one visits (other than occasional battery maint). Cheaper cooling in the data center, data center is somewhat more habitable, etc.

Google 12VDC proposal better.

[Animats]

There's no particular reason that 380 VDC distribution should help efficiency. You still need about two more levels of switching power supply before power reaches the ICs.

Google's proposal that motherboards should need only 12VDC made more sense. Drives already run on 12VDC, and there's already a level of power conversion near the CPU to get the desired CPU voltage. The USB devices do need +5, but a 12VDC to 5VDC switching converter can handle that. And single-voltage power supplies are more efficient and simpler than multi-voltage ones.

Re:Google 12VDC proposal better.

[Mr Z]

The current carrying capacity of the wires would need to be about 30 times larger, though, to deliver the same amount of power. That's pretty huge. To go to 12v everywhere, you'd need huge current-carrying wires everywhere (think "as big as your car battery cables or bigger"). To carry 1kW through a 380V line, you only need to handle 2.6A. To carry 1kW through a 12v line, you need to handle 83A. And that's just one beefy server.

Now think of your house wiring. Outside of your major appliances, where do you see runs higher than 15A or maybe 30A? There's a reason high voltage is good.

Re:Google 12VDC proposal better.

[vlm]

The original low speed USB electrical spec was pretty much classical 5 volt NRZI TTL. So insisting on 12 volt supply would mean every USB device would require a 5 volt regulator inside it to talk to the data lines, and the data lines would need protection circuitry on both momma boards and all USB devices because TTL traditionally gets really pissed off when an input voltage rises about its power voltage in case of a short. CMOS gets pissed off too at over voltage. It would just be a bad scene.

Something like RS-485 but really faster would have been "better", but ...

Re:Google 12VDC proposal better.

[hpa]

USB is designed to enable inexpensive devices. 5 V is so that when cable losses are counted in, you can use a LDO linear regulator to obtain 3.3 V (Vcc) without excessive losses. 12 V would require a switching regulator.

Re:Google 12VDC proposal better.

[Wonko the Sane]

what about salt water filled copper pipes for the conductors?

Excellent idea! Hydrogen gas, oxygen gas, chlorine and an ignition source all in the same package. What could possibly go wrong?

why 380v?

[wierd_w]

Wouldn't it make more sense to drive at 12v with an insane amperage behind it, than to drive at 380v and garantee the necessity of a voltage regulator rated for high voltages?

I mean, the whole reason for doing away with ac current was to eliminate the rectifier and regulator circuits, which belch heat into the data center. Using 380v, which no datacenter device that I know of uses natively (well, maybe the innards of a crt, but that's actually much higher than 380v... AND a deadend tech.), seems kinda... well.... unproductive.

Is it because of impedence problems or something?

Re:why 380v?

[Urban Garlic]

Basic Ohm's law -- the resistive loss through a DC wire is the voltage drop across the wire, times the current through the wire. But the voltage drop across the wire is proportional to the current, it's just I*R, so the total power dissipated in the wire itself (i.e. not transferred to the load) is I*I*R. So, you want the current going to the load to be as small as possible. But, of course, the load still needs to get all the power it needs, so the operating voltage (which is distinct from the through-the-wire voltage *drop*, of course) needs to be higher if the current is lower.

So, high operating voltages reduce distribution losses.

The same analysis works for AC too, and is the reason that trans-continental transmission wires have such crazy-high voltages. AC has additional losses due to radiation and induction, of course.

Re:Why 380v?

[RichMan]

440 * sin(120) = 381.05 ....

3 phase has 2 ways of looking at the voltages, Y or delta.
The 3 phase delta is 440v when you measure between any pair of the 3 wires. The center point is ground. You don't see that in delta, but you do when measuring it in Y form. The same signals that are 440v when measured as a pair are 3 x 380v when looked at in the Y configuration.

So 3 phase 440v gives you 3x 380v to ground.

As to the 12v/5v/1.5v/ whatever you are going to have to do DC to DC all over the place. Better to have as high a voltage as possible for less current and less losses.

You'd need much larger conductors

[bigtrike]

Lower voltages require larger conductors to carry the same current. Copper isn't that cheap.

Re:You'd need much larger conductors

[nzac]

No the skin effect is for (High frequency) AC.

For DC impedance is determined by the material and the cross section area.

It does make the cables easier to bend though.

We have 48VDC as one standard...

[mlts]

If one has worked in a telco, we already have a standard, and that is 48VDC. This is the domain of the Sun Netras of yore.

If I were to recommend a voltage, why not plain old 12VDC? Yes, the amps have to be high, but we already have a connector for this (beats wiring up things by hand and throwing a breaker), and it is not hard to find off the shelf hardware to support this, be it batteries, power distribution units, inverters/converters, solar panels with MPPT controllers, and so on. We have two large markets (RV/marine) that are dedicated to 12VDC.

Why not just use an established standard? 12VDC works and has a lot of support, or if a higher voltage is needed, then 48VDC.

384VDC just seems to be asking for trouble. It would require yet another separate connector that can't be plugged into 120VAC or 240VAC, generators would have to have an adapter for it. It would require a complete retooling to get to that standard.

Making another voltage level is throwing the baby out with the bathwater. Why not just go with an established DC voltage level?

Take 12VDC. Most generators, from the expensive inverters by Honda or Yamaha can generate that, as well as the construction grade open-framed ones.

Re:We have 48VDC as one standard...

[Wonko the Sane]

Yes, the current will be higher at lower voltage. This does NOT correlate to needing thicker wires, as the wire has to withstand not current but power which is the result of multiplying voltage with current.

You've managed to be right while also being wrong at the same time.

You could use voltage*current to calculate the thermal losses in a conductor but what you've done incorrectly is assume that "voltage" in this equation is the voltage between the conductor and ground.

The correct way to calculate losses in a conductor is current * end-to-end voltage difference

The end-to-end voltage difference is directly proportional to the current so the most efficient way to calculate the losses is current squared times resistance.

Since the surface area of a wire is proportional to the square of the wire diameter and the conductivity required is proportional to the square of the current carried it ends up that wire diameter is directly proportional to the current.

Re:And in related news...

[DickBreath]

Would an AC current be generated by Tesla spinning in his grave?

Re:Power monitoring

[blair1q]

Hall effect.

In the presence of a static magnetic field (as around a conductor carrying a constant current), electrons in the clamp circuit, which also carries a DC current, will be pushed to one side of the clamp conductor, inducing a voltage relative to the other side. Measure the voltage and you know the current in the wire it's clamped around.

My GTI's steering is wonderful.

[zerofoo]

My 2008 VW GTI has an electric assist rack, and it is one of the best systems I've ever felt in terms of feedback and heft. I've also driven overboosted hydraulic systems that feel like mush.

Electric assist steering can be done well, and hydraulics can be done poorly. The technology isn't to blame, it's the engineering that matters.

Telco power connectors

[DragonHawk]

Only very stupid engineers design power connectors that can fit both ways.

The DC power supply connections in telco equipment is generally screw terminals and spade connectors.

Re:Plus brushless motors

[LinuxIsGarbage]

Computer fan motors are brushless DC. But really they are permanent magnet AC motors with a simple VFD (variable frequency drive) in it. I suspect RC planes are the same.

In the industrial world VFDs are very popular. On anything from a 2HP conveyor to a 1000HP+ piece of equipment. They rectify three phase input to DC, then convert it back to AC at the desired speed. Some are setup so you can have one central rectifier, and multiple inverter sections for your different loads.