"Heat Wheel" Could Lower Data Center Power Bills 97
miller60 writes "An air conditioning technology called the 'heat wheel' is getting a test drive in data centers, and early adopters cite impressive reductions in their power bills. The heat wheel — also known as a rotary heat exchanger or Kyoto Cooling — is a refinement of cooling systems using outside air. Rather than introducing exterior air directly into the server room (the air economization we discussed recently), the heat wheel briefly mixes the outside air and exhaust air to create an air-to-air heat exchanger. A data center in the Netherlands using this approach only has to use chillers 11 days a year." The article points out that the heat wheel is not new, but it hasn't been applied to data centers until recently.
how about dropping the ac - dc - ac - dc to one AC (Score:2, Insightful)
how about dropping the ac - dc - ac - dc to one AC - DC part?
Re: (Score:1, Offtopic)
Re: (Score:2, Insightful)
Re:how about dropping the ac - dc - ac - dc to one (Score:4, Interesting)
It's easier to control opening and closing of transistors and diodes and other components associated with digital signals that will be present in any IC circuit with DC.
If AC was used, you would have to put the ability to control the signals and such into each and every circuit which would cause an excessive requirement for materials and components as well way more heat if it was even possible. This is why the power supply breaks this down and give DC in the various voltages where it is needed. Using DC also allows you to create a base signal that is increased or decreased in order to have a desired effect on a component that simply would be possible with an on off switch like AC.
My explanation probably sounds strange and I could be completely off on a few of the generalities that I purposely attempted to keep general. When dealing with computers and ICs, you really need to look at the operations more like a radio station broadcasting then switches turning on and off. It is the words in the broadcast (the peaks and valleys of the radio waves) that turn other gates on and off and allow the computational happiness to happen. The communication built into a digital circuit is a lot like a broadcast with a timer to declare a signal length for on and off when the base wave is changed. AC is just the most available source of power so it is commonly used.
As for 12 volt systems like in cars, the voltage is really too low to be useful for long. The drain from a 450 Wat power supply on a 12 volt system would be something like 37-38 amps which means a lot of heat will be generated when delivering the power where in a 120 volt AC system, it is something like 3-4 amps which doesn't generate as much heat or require as big of lines. Now I know a 450 won't be pulling 450 but it gives an idea of the requirements and hopefully helps answer your question/statement.
Re: (Score:2)
Your question is a little ambiguous because any voltage that isn't constant (DC) can be called an AC signal. There's nothing special about the sinusoidal voltage you get from a wall socket, it's just not much use to a computer. Digital systems are supposed to act like boolean logic, so ideally there are only two voltage levels (0V and 5V, say) allowed in the circuit. They do have AC (time-varying) signals -- a clock, inputs, and outputs. Transistors in the circuit are switched on and off in response to the
Re:how about dropping the ac - dc - ac - dc to one (Score:5, Informative)
There are seemingly not many fans of the DC powered data center on /.
Every little bit helps and point of load DC-DC converters are quite efficient, thus do not generate much heat. Additionally, since the back-up power for a data center is based on batteries... well, you can do the math on that. Generators are a different issue, but even they don't have to be AC, though probably more efficient if they are.
Every reduction in heat generation improved energy efficiency. Likewise, running on DC would reduce energy consumption by some measure. There is a reason that telephone exchanges are run on -48VDC, and it's not some fscked up reason like "oh, that's how they ran the first switches in England, and we never got around to changing."
It will take many small steps to achieve big results. DC power is but one of them.
Re:how about dropping the ac - dc - ac - dc to one (Score:5, Interesting)
There is a reason that telephone exchanges are run on -48VDC, and it's not some fscked up reason like "oh, that's how they ran the first switches in England, and we never got around to changing."
Ah, I know that one, or at least half of it. The reason for the negative voltage is electrolysis [wikipedia.org]. A positive voltage would result in a migration of metal from wires exposed to the environment (telephone poles) to earth. Negative voltage makes the infrastructure last longer.
As for the magnitude being 48 Volts (actual spec. usually 36-72 volts) it most likely has to do with the maximum voltage drop between the central office and the terminal (phone).
Re: (Score:2)
You are quite correct but the point is that using DC powered equipment still has more advantages than AC powered equipment, especially in that 24/7 environment. The 48 volt setting is also due to the fact that your POTS line was powered by it. There are safety factors and equipment requirements from long ago that help determine such parameters. The DC power to the phone is ages old, and ensured that your phone worked even when your power was out. Early technology required this, reliability helped ensure it'
Re: (Score:2)
The great advantage of AC over DC was the ease of voltage conversion. Nowadays, DC to DC conversion is almost as easy (though except at small currents, it's still done by converting to AC and then back to DC. The DC->AC stage is what has been improved). But of course there's an absolutely enormous AC infrastructure, and generally no compelling reas
Re: (Score:3, Informative)
Switching power supplies are easy to understand, here is the basic idea:
take a bunch of capacitors and charge them in parallel (for a voltage increase) or in series (for a voltage drop) and then discharge them in the opposite configuration.
So if you charge two caps in series at 3V when you discharge them in parallel they will produce 1.5v each.
Think of it in the same way as batteries, in series they voltage is added together. This is why it is called
Re: (Score:2)
It converts the power to DC, then to high frequency AC, then runs it through a transformer, then back to DC.
That's a charge pump, not your typical switching power supply. Check your
Re: (Score:2)
It's 48 volts because 50 volts and above you need an electrician's license to work with it, and the people who originally started using the standard (phone companies, I believe) didn't want to pay union rates for an electrician.
Re: (Score:2)
And every little bit of extra cost adds up... And make no mistake, a DC datacenter costs a LOT.
No they aren't. There's nothing magic about DC that makes it more efficient to convert. Why do so many people seem to think that when the input and output voltages are a bit closer together, it's more efficient? In fact the opposite is closer to accurate.
AC can be converted to DC trivially, with bare mi
Re: (Score:2)
DC Power for data centers is a good technology, but if you are suggesting 48VDC you must own copper futures! Most of the data center scale systems proposed today are 600VDC (nominally), as you can easily get to 6MW on a single bus. 48VDC would only give you 300kW, assuming your runs were very short. (Bus size based on a 4-wire, 5,000A bus duct with two phase busses per pole.)
Re:how about dropping the ac - dc - ac - dc to one (Score:4, Interesting)
That's pretty normal in "teleco" equipment. 48V is standard for exchanges etc, and many server manufacturers provide it. It definitely helps for some circumstances and makes battery backup easier (generators, however, are disadvantaged since they need to be on the other side of your rectifier)
You always end up with a fair amount of invertors for all sorts of stupid things that you have to get AC for (e.g. service engineers laptop power supplies). You also end up with lots of big copper cables and / or buzz bars. That gets quite expensive. I've seen whole buildings kitted out for that, but it needs real pre-planning.
Re: (Score:3, Informative)
Laptop power supplies are separate from the laptops. All you need is access to the appropriate DC feed and a correctly sized adaptor. "appropriate DC feed" is often 12V, and usually one of only a few standard voltages.
Re: (Score:1)
Most laptop power bricks put out 20V
Re: (Score:1)
Generally an emergency is not the time for messing around trying to work out laptop power supply compatibility (and any servicing you do during a power outage is likely to be emergency related).
Your service engineer, normally from a random vendor, arrives from outside. He has never seen your system before. You give him a power interface he understands and has used before. That means AC.
Re: (Score:1)
how about dropping the ac - dc - ac - dc to one AC - DC part?
The Offline UPS [wikipedia.org] has a possibility of not kicking in soon enough.
If your data centre guarantees four nines [wikipedia.org] uptime to clients...
Re: (Score:2)
I sure am glad someone brought this up.
I don't know how many headaches I have had to deal with because someone thinks a $50 ups at office max is the same thing as the Double-conversion ups that cost quite a bit more. Ad to that someone who thinks a rinky dink 5500 Wat portable generator can keep the computers up and running during a power failure and we have some serious problems with brownouts and so on.
Re: (Score:1)
Not until superconductors are workable. Even in the short runs between servers, DC tends to have higher losses. You're better off spending money on higher efficiency power supplies.
http://event.on24.com/event/95/75/4/rt/1/documents/player_docanchr_1/wp63_fr.pdf [on24.com]
Re:how about dropping the ac - dc - ac - dc to one (Score:5, Informative)
I would suggest you review the information in the paper you linked regarding HV DC distribution. They show it to be significantly (for the values that pass for sig. in this case) more efficient than conventional AC power distribution, and help simplify the PSU design. In addition, I do believe that use of higher voltages would, for computer PSUs, at least, allow for more efficient DC-DC designs, a fact not accounted for in the paper. (Unless I messed something, they only considered the removal of the PFC component of the PSU.)
HV DC is a rather clear winner. The only reason AC is better than DC in this environment is because it's at an inherently higher voltage. Obviously 48V DC is going to have significantly higher ohmic losses because it's going to have to carry twice the current to deliver the same power, while still needing to going trough a DC-DC anyway. I frankly can't imagine why 48V was seriously proposed.
As an aside, I do believe that, for a given voltage and power (rms values for AC) DC has exactly the same ohmic losses as AC. (Less, if you count the skin effect.) The only reason AC won the "War of Currents" is because it could be distributed at high voltages and stopped down at its destination. (FWIW, I think that DC would have lower losses than AC in the ultra high voltage transmission lines, as corona discharge actually dominates losses at that point, and is largely a function of peak voltage, though I've not done any research specific to this topic.)
Re: (Score:3, Informative)
I would suggest you review the information in the paper you linked regarding HV DC distribution. They show it to be significantly (for the values that pass for sig. in this case) more efficient than conventional AC power distribution, and help simplify the PSU design.
But it also says the "high voltage ac" option which appears to the the european 230/400V three phase system is almost as good and that is a system that should be easilly deployable in datacenters worldwide without needing any special kit (PC po
Re: (Score:2)
That paper was written by the CTO of APC. And yet, APC won't actually manufacture any UPS or surge protection devices that function correctly on the 240V single phase electrical system in North America. Lots of people could change even their small home computer systems, as well as the computer rooms of small businesses (a great many of which do not get any three phase power, and most that do get less than 240V) in the USA. Their competition won't manufacture these products, either. Chicken. Egg. I gue
Re: (Score:2)
This difference requires a different UPS design if a bypass relay is included
Afaict in most of europe there is no gaurantee which wire is live and which wire is neutral so a UPS designed for european use would have to be able to tolerate either side of it's input being live, given this I would not expect there to be a problem running it on american 240V
Re: (Score:1)
Many plugs for low power however have just two wires in the free-socket end of the power cord (usually an "8" connector) and these of course cannot be relied upon.
Re: (Score:2)
Which wire is hot is well defined for three-pin plugs
That may be true for some types but it is not true in the general case. German plugs use side clips for earth and are not polarised. Italian plugs use three pins in a straight line and again aren't polarised. French plugs are polarised but from what I can gather the french aren't very carefull about which way round they wire them.
So any UPS sold for use accross europe will have to be able to tolerate the incoming live and neutral either way arround. Given
Re: (Score:1)
The "Euro" plug that is now standard in the EU (except for UK) is polarized and will only mate in one orientation.
It will still be some time however before these are used everywhere. There are still a lot of legacy three-pin or two pin sockets around and most of these are not polarized.
Of course if a plug or socket is miswired, there is no way to get around that. Most Euro cables for appliances are molded and do not require the customer to wire it. Socket wiring is down to th
Re: (Score:2)
The "Euro" plug that is now standard in the EU (except for UK) is polarized and will only mate in one orientation.
If you mean a CEE 7/7 plug which is as far as I can tell the most common earthed plug in europe nowadays it is polarised when used with a french socket but unpolarised when used with a german socket.
If you mean some other plug type please specify exactly which plug type you mean (link to a picture if needed).
Re:how about dropping the ac - dc - ac - dc to one (Score:5, Informative)
DC at these higher voltages (340V to 500V) is more difficult to work with than the corresponding AC. As a result, the initial installation costs, and some maintenance costs, are higher. For example, a circuit breaker that can safely shut off a short circuit fault current at 120V AC is more limited at DC. The maximum voltage they are rated for in DC is 48V. Breakers for higher AC voltages can do relatively higher DC voltages. But you'll need breakers of a class for well over 1000V AC in order for them to be able to handle 500V DC. It is harder to extinguish an arc in DC because there is no zero crossover times that happen at 100 or 120 times a second for AC.
The higher voltage AC needs to be considered at least as a reference point. Then if the savings the higher voltage DC offers (such as simpler PSUs) can exceed the extra costs involved of HVDC distribution, it could be viable. Otherwise I suggest someone putting in a large data center in North America special order their power at the 416Y/240 service voltage (uncommon, but doable with 3x 240V transformers), and just plug each computer into a 240V circuit (1 hot, 1 neutral, 1 ground) much like already done in Europe (except in North America it will be 60 Hz which is irrelevant to switching mode PSUs).
Re: (Score:1)
I don't get it (Score:2)
Re:I don't get it (Score:5, Informative)
Re:I don't get it (Score:5, Insightful)
Because outside air contains all sorts of things that you don't want in your data center: humidity, contamination (dust, pollen), etc. While you could get rid of dust with filters (that would need frequent replacing), there's no simple answer for humidity.
Humidity (Score:2)
"there's no simple answer for humidity."
Move somewhere where the climate is drier and cooler.
The outside air here currently has a dewpoint of -6c and the temperature is -2C
Re: (Score:2)
I would guess that it's better then directly venting outside air in and exhausting inside air out due to humidity levels. The inside air is already conditioned, but it has unwanted head. By mixing it with controlled amounts of outside air you can lower the temperature without impacting humidity as much. If you were to bring in outside air directly, you'd have to dehumidify it as well.
That's just my guess though.
Right, but... (Score:1)
My car has an air-to-air intercooler that doesn't mix the air streams but very effectively exchanges the heat. All you really need is a fan to keep the air moving and you're done, no silly wheels or compressors involved.
So it's a heat exchanger? (Score:4, Insightful)
!news. Many (most?) well designed AC systems employ heat exchangers.
What's the diff between a rotating and a conventional heat exchanger? Efficiency? Cost? Of course TFA doesn't mention any of it.
Re:So it's a heat exchanger? (Score:4, Informative)
!news. Many (most?) well designed AC systems employ heat exchangers. What's the diff between a rotating and a conventional heat exchanger? Efficiency? Cost? Of course TFA doesn't mention any of it.
That happens a lot when you don't read TFA:
"This system has all the benefits of Airside Economizing, without the exposures of airside economizing like contamination and humidity control."
Re:So it's a heat exchanger? (Score:5, Informative)
He may well have read TFA. The article and your post only mention 'Airside Economizing' which mixes the two air flows. Heat exchangers don't necessarily do that, and have been around for a very long time. Even the old VW bug used a system that didn't mix. All this system does different is use a wheel, adding a moving part to a very common method of stripping waste heat from plant exhaust.
The article specifically omits comparing it to other heat exchanger systems. They give an example data center, but don't tell us what the data center did before.
Most telling is the quote: "Heat wheels have been used for many years in industrial air conditioning, but never in data centers." Other than asking "really?" and "why not?", since data centers are industrial use, the question is why aren't they comparing like with like here? There were plenty of things like this when I was dealing with 'super-insulated' building construction in the 80s.
This article is really just the manufacturer's PR release regurgitated, and is properly tagged 'slashvertisement' here. Data center managers should definitely know about heat exchanger technology, but get a better source.
Re: (Score:3, Interesting)
A heat wheel is an odd way to cool a data center; it requires fairly low outside air temperatures to work. We've had better luck with indirect evaprative coolers pre-cooling the hot-aisle return air and discharging to the space. Unless your wet-bulb temperature is below 60F, you will still need some compressor cooling, but since wet bulb temperature is never higher than dry bulb you get much better range and efficiency.
Re: (Score:3, Insightful)
But you still get stupid replies from others who don't understand what they read.
A heat exchanger is NOT airside economizing. In fact, the two are polar opposites, though I suppose there could be some manufacturers out there that conflate the two in their ads.
Re:So it's a heat exchanger? (Score:5, Informative)
Looks like the proposal is to use rotary regenerative air heaters [wikipedia.org] (often known as Ljungstrom(tm) heaters in the power plant biz) for a low-temperature application.
I once gave a presentation to senior management on a situation with the air heaters at our plant; I had to practice saying "Ljungstrom re-gen-er-a-tive heater" for two weeks before I could do it without stumbling!
sPh
Isn't the explanation completely wrong? (Score:2, Interesting)
How does this thing really work? It seems to me to be carrying the heat round on the wheel from one air mass to the other rather than mixing the air? If you RTFA it seems to imply that the first mix and then unmix the air. That would be worth of a patent...
(moderate: -1 troll - suggests Reading TFA anon-comments: Ewe muzt be newzor here)
Re: (Score:3, Interesting)
To me it looks like this:
The wheel allows air to pass through it, but is very conductive - it quickly changes temperature to match the ambient temperature. As it is rotating, a section heats up when it is exposed to the inside air, then cools when it is exposed to the outside air. This in turn cools the inside while heating the outside. The air itself doesn't actually mix.
Of course, this could be completely wrong, but that's what it appears like.
Re: (Score:2, Interesting)
Re: (Score:1)
Re: (Score:1)
Why not just stick a big heat sink with fins all over it in there. The metal could transfer the heat from the inside air to the outside air. There must be some reason why rotating would make it better.
Although metal is an excellent conductor of heat, if air is not moving past it, the whole wheel will reach thermal equilibrium -- very quickly, since metal is such an excellent conductor. Mind you, I'm not saying it will reach a uniform temperature, it will be cooler on the side exposed to cooler air and warmer on the side of the warm computer hardware, but the distribution of heat throughout the mass will reach a steady state.
The article says they mix the air, which you wouldn't have to do with a solid metal heatsink heat exchanger.
I think that may have been an exaggeration. As has been noted, this was a Slash
Re: (Score:2)
> That would be worth of a patent..
You are late by about 100 years. Look up Ljungstrom(tm)...
sPh
Re: (Score:1)
Well, several of seem to be implemented with heat wheels, but, for example Alstom makes it pretty clear that one of the benefits of their system is that the gasses don't mix. So I'm not sure how that's relevant?
Re: (Score:2)
Re:air exchanger (Score:4, Informative)
That works only if the humidity and dust are taken care of. Humidification isn't too hard, but de-humidification takes nearly as much energy as air conditioning.
A heat exchanger doesn't take much power at all to run, quite probably than a filter + humidity control (keep in mind, a filter costs power because the blower has to work against it's resistance).
Re: (Score:2, Informative)
The computer room Lieberts commonly used in data centers actually cool, then re-heat when the temperature is within range, but the humidity is too high.
Re:air exchanger (Score:5, Informative)
Actually, it is. Not only are all humidification plants natural growing ground for fungi - which, if you think about it, is a really bad thing since all the spores will be blown straight to the building - but it actually takes a lot of water to humidify large quantities of air.
A heat exchanger of the type described (which, BTW, doesn't work by mixing air, it works by using the mass of the wheel as a heat battery and moving it between the two airstreams) indeed requires very little power to run; just enough to keep the wheel turning against friction losses. Here in Finland we use the same system for, ironically enough, to cut down heating costs. I once saw the ventilation plant for a large school building; the wheel was just a meter in diameter and moved a hundred rotations or so a minute.
Slashvertisement (Score:3, Insightful)
Why not just say that data centers are using heat exchangers and outside air to cool their computer rooms.
All the stupid wheel is is a heat exchanger like any other. Many types of heat exchangers allow the inclusion of outside air, though I would think it would be better to keep your computer room air closed from the outside if it is possible.
Re: (Score:1)
Re: (Score:2)
Shame the article didn't point any of that out.
Here's the "rest of the story" (Score:4, Insightful)
you'll like this.
your pal,
kdawson
sales division,
slashvertising inc.
Enthalpy wheel? (Score:4, Informative)
This looks exactly like an Enthalpy Wheel, which transfers some of the moisture as well as heat between the exhaust and intake streams.
I am actually quite shocked these things have NOT been used for data center ventilation before. The bigger the ventilation job, the more these things make sense.
Only thing I can think of, though, is that data centers probably don't have high ventilation requirements... machines don't need a constant supply of fresh air for breathing, so a lot of it can be cooled and recycled.
But if all they're doing is transferring heat (and not humidity as well) then there are better options available.
=Smidge=
Re: (Score:3, Informative)
Lol... netherlands (Score:5, Insightful)
"A data center in the Netherlands using this approach only has to use chillers 11 days a year."
Umm.. yeah... the netherlands is generally a cold place. Not really saying much if the listener knows a little something about geography and weather.
Re: (Score:1)
Re: (Score:2)
Which you get about 5 days a year.
Heck even in the slightly warmer UK where we don't get the same extremes due to the gulf stream, the datacentre near us its AC about 2 days a year. We do this terribly technical thing cslled "opening the window" which is much cheaper than AC.
It's actually harder to keep these places warm than cool, even with all the machines in them.
Re: (Score:1)
The Netherlands has a maritime climate, with cool summers and mild winters. The average temperature is 2C (36F) in January and 19C (66F) in July, with an annual average of about 10C (50F). Clouds generally appear every day, and in the winter months fog often abounds, while rainfall occurs frequently. Average annual rainfall is about 76.5 cm (30 in). The mild, damp climate is ideal for dairying and livestock raising, but the limited sunshine restricts the growing of food crops . [nationsencyclopedia.com]
Yeah, 11 hot days in July. Well, I suppose there is a good reason not to build your data center in Dallas 28 days in July above 90F [cityrating.com].
Re: (Score:3)
Re: (Score:2)
So, this is a good argument for moving high demand data centers to colder places. I know that this means that latency takes a hit when the data is FAR from you. For some applications, this wouldn't matter, latency wouldn't make a huge difference. I'm thinking sites that display lots of static content, like bank web pages.
Re: (Score:3, Funny)
Re: (Score:2)
Quote: " ... "A data center in the Netherlands using this approach only has to use chillers 11 days a year."
Reply: Umm.. yeah... the netherlands is generally a cold place. Not really saying much if the listener knows a little something about geography and weather. ..."
I can't believe this reply to an actually informative post is modded 5/insightful.
So, by this reply's logic, a blowtorch in the Netherlands is colder than a blowtorch in Arizona, thus the lesser need to cool the sealed, temperature and humidit
ERV (Score:4, Informative)
Energy Recovery Ventilation Systems...
Been around for years at least 15 or better that I know of. I used to work
for a company that built these units.
Re: (Score:2)
not news. ancient (Score:1)
Are these people stupid? (Score:5, Informative)
Are these people stupid?
Heat wheels, free cooling, ground loop heat pumps, these are all technologies that have been around for 50 years, and have been mainstream for just as long.
I work in the HVAC controls industry, and even the smallest private schools and the like use heat recovery wheels, free cooling, and many use ground loop heat pumps.
Goodness, every packaged rooftop unit manufacturer has done free cooling for decades. They are usually controlled by enthalpy sensors or return air CO2 levels. No fancy computers are required to control them, although our job is to add that functionality.
Of all industries, the IT cooling one seems to be in the dark ages.
And yes I know they haven't been using these technologies, I work around server rooms quite a bit and can't get over how much energy they waste by not economizing.
It's not rocket science and it should be blatantly obvious to the engineers that design the systems.
Re:Are these people stupid? (Score:4, Informative)
Now that residential forced-air furnaces have fairly high efficiencies, the next step was to introduce this technology into homes. Small-scale air preheaters have been required by new construction building codes for many years now in some parts of the country, like Minnesota.
This data center cooling system is just the same thing with a transposed air flow - nothing radical, but there was likely never really a practical use for medium-scale units until recently.
Re: (Score:1)
Obsolete tech... (Score:2)
OK, I'll be the first to say it. Who cares?
Seriously, think beyond the room itself. Computers capable of complete immersion in non-conductive liquids, GPUs offloading CPU processing thereby spreading heat "signatures" around the board even further, we're likely less than five years from a complete revamp in the way we cool computers themselves, thereby quite possibly eliminating this kind of cooling requirement altogether from a Data Center.
Re: (Score:2)
Oh, like computers from the 80s? ;P
http://en.wikipedia.org/wiki/Cray-2 [wikipedia.org]
The Cray-2 was immersed in liquid.
Bad news (Score:1)
A data center in the Netherlands using this approach only has to use chillers 11 days a year.
This came as a shock to the hard working serial chillers.
steel shops used this (Score:2)
My dad, a tool and die maker, worked in a building that was build in the 40-50's. It had something that worked on the same principle. There was this long duct shaft that went up the east wall, across the room, and down the west wall. Where the ducts reached the floor there were huge fans, two at each entrance. In the morning the east fans would run. In the evening the west fans would run. During the winter they would change the direction the fans pushed air.
That's not how they work (Score:2)
So....what you're telling me is (Score:2)
That they're opening the window and letting the heat out?
Maybe attaching a fan to help move the air?
Brilliant!
Re: (Score:1)