Microturbines Power, Cool Servers Simultaneously

jfruhlinger writes "The infrastructure of a large data center poses two main problems: You need to find a way to reliably power all those servers, and you need to figure out a way to deal with the heat those servers put off. Syracuse University and the University of Toledo are experimenting with one gadget to solve both problems. Small power units that run on natural gas, called microturbines, provide reliable DC power separate from the utility grid, and their heat output can paradoxically be harnessed to cool the servers and transmit the heat to other buildings on campus."

Purdue University has this

[Maximum Prophet]

Purdue has done this for years, but with macro turbines. The main physical plant provides power, chilled water and heat most of the University.

Re:

[JoeMerchant]

Of course, the percentage of energy recovered is laughably low, but at least they are getting something back.

Re:Purdue University has this

[goofy183]

Citation?

Many EDUs and other large business campuses use tri-gen plants and from everything I've seen they arguably are significantly more efficient per unit of input engery: http://en.wikipedia.org/wiki/Trigeneration [wikipedia.org]

Re:

[Anonymous Coward]

Actually, the main purpose of the power plant is to provide heating and cooling to campus. Since it utilizes what would otherwise be waste heat, it has very high efficiency.

Can they sell unused power back to the grid?

[Joe_Dragon]

Can they sell unused power back to the grid?

Re:

[dnwq]

In this server farm we obey the second law of thermodynamics!

Re:

[mosb1000]

It'd be cheaper not to use it in the first place.

Re:

[Anonymous Coward]

Why would they? A "micro" turbine is surely less efficient than whatever large-scale system the power company is using.

In other words, the gas they use to make the electricity probably costs more than the power company will pay them for it. So it would be a net loss to sell back to the power company.

Re:

[wagnerrp]

For shaft power, the large scale turbine running at higher pressures and temperatures, with fancy intercooling and heat exchangers, is going to provide much higher efficiency than a small one. If the waste heat from the small scale turbines is used to run chillers or boilers, the total thermal efficiency of the whole system can be higher than a larger shaft-power-only turbine. Of course, there is nothing preventing the larger turbine from being hooked into a similar combined cycle system as well.

Re:

[wagnerrp]

That sounds wrong. Standard utility rates in North America are somewhere around $0.07-$0.15/kWh, not the $0.005/kWh you're claiming.

Re:

[benjamindees]

there is nothing preventing the larger turbine from being hooked into a similar combined cycle system as well.

There is the economic fact that people who can make use of low-level waste heat don't live next to power plants.

Re:

[wagnerrp]

In college, I lived three blocks from a natural gas fired tri-generation steam plant rated at 25MW electric, 70MW thermal, and 18000 tons of refrigeration. If you weren't paying attention, you wouldn't even know it was a power plant.

Re:

[MobyDisk]

The last 30 seconds of the video specifically mention that.

Re:

[sam0737]

The TFV (video) says yes.

Re:

[Patch86]

Presumably yes. But seeing as it's a gas-powered turbine, it'd probably make more sense to just not generate more power than is needed.

Awwww shit.....

[Immostlyharmless]

My gaming PC already sounds like a Hoover and now you're telling me the next evolution in cooling is to put a turbine in it? :o(

Re:

[Culture20]

My gaming PC already sounds like a Hoover and now you're telling me the next evolution in cooling is to put a turbine in it? :o(

Yeah, and it runs on natural gas, so there's that extra noise too. Stock up on frozen burritos.

Re:Awwww shit.....

[skids]

No this is for server rooms. Something tells me most people won't want to run a line from their natgas system into their server.

Speaking of server noise, though, I've often wondered if a laminar-friction impeller might at least not have that high-pitched whine. Basically this is a squat horn-shaped surface spun really fast. The air enters through the hole and gets accelerated by laminar friction out in all directions -- so it would have to be redirected with a hood to produce a lateral flow compatible with server fans, but then might be able to "entrain" like that ridiculous looking Coand-effect donut fan that Dyson sells. The main problem is the bearing has to fit around the big hole, so that's much more bearing adding to the cost of the unit. Though it might be possible with careful motor design to make the whole plate levitate rather than ride a bearing. The huge advantage, other than the lack of turbulence, would be there's no leading surface on which dust and debris can perch.

Re:

[skids]

The idea is that the device is completely symmetrical with no blades or veins. There may be some imperfections, but they would produce extremely low amplitude signal. Minus the actual noise of the airflow itself, should be as quiet as a hard drive.

fanless spinning heatsink

[Chirs]

http://www.extremetech.com/extreme/89710-the-fanless-spinning-heatsink-the-heatsink-is-the-fan [extremetech.com]

Re:

[skids]

Not exactly what I had in mind, but very, very interesting. Thanks for the link!

Turbo button?

[HornWumpus]

Will it finally have meaning?

Will future PCs suffer from turbine lag?

Re:

[squidflakes]

Oh man, I would have gotten that headshot but my turbine was spooling down.

Re:

[Jeng]

Well it has the little red button to make it non-functional, but I can't locate the clutch.

http://www.msi.com/product/vga/NX8600GT-Twin-Turbo.html [msi.com]

Re:

[GameboyRMH]

Ball-bearing variable geometry turbos means little lag no matter what you do with the clutch.

Re:

[icebraining]

From here? [mini-box.com]

Re:

[wagnerrp]

He said "for a server". While the PicoPSUs are fine for a desktop with modest graphics, they're not going to do much for a multiprocessor server. Besides which, 12V at 50+A is some nasty stuff. You would be better with telco grade -48V hardware if you have to pass a significant amount of power.

most telcos use -48V DC

[Chirs]

So carrier-grade gear will be DC-DC.

Re:

[wagnerrp]

Traditional gas turbine generators run 5-10MW for small ones, up to 100MW for large ones. In this case, a microturbine would be something small on 10-100kW. Note that a large automotive turbocharger is capable of 100kW or more. On the plus side, multiple smaller units means a more redundant system with less overhead. On the minus side, several small turbines is both more expensive and less efficient than one large one of the same output. It would be cheaper to just go with two large turbines, each capa

Re:

[afidel]

Toledo only needs 260 kW electric and 100 tons of cooling, much too small for a traditional gas turbine. In fact BHP's largest project [bhpenergy.com] is 2MW electric. I know I've lookup into such a setup but the noise level was unacceptable as our generators sit in the middle of the three buildings in our campus.

Re:

[wagnerrp]

Toledo only needs 260 kW electric and 100 tons of cooling, much too small for a traditional gas turbine.

Yikes. We've got that much power run to our little compute cluster servicing half a dozen graduate students. We've got a pair of 400A 2-phase 220V panels for our computers, and about as much running to a pair of AC units. Under full load with everything running, we might hit half that. The combustion lab at the other end of the building runs a pair of 800A 3-phase 480V service lines to power resistance heaters for their high pressure air tanks. The lab back on main campus uses around 2MW of steam from

Re:

[afidel]

Hehe, I run an S&P 500 company on 40kW and 12 tons of AC, virtualization and SSD's for high IOPS loads are a wonderful thing. In fact I'll probably be well below 30kW in 6 months if I can get the 4 shelves of 450GB drives to fill out my new SAN (damn floods in Asia) so I can retire the old one and all the old P4 based servers connected to it that constitute almost half my physical server count at this point.

Re:

[wagnerrp]

Admittedly, "full load with everything running" includes several hundred each of P4s, single core AMD64s, dual core AMD64s, Core2Duos, and a smattering of i7s. The older single core boxes are rarely even turned on any longer.

chp

[thejaq]

This is combined heat and power many facilities do it. It is green in the sense that energy is conserved because waste heat is used rather than discarded. A data center seems to be a good opportunity. The turbine converts 1 CH4 unit to 0.3 electricity, while the absorption chiller will move about as much energy as it consumes (COP 1), which means the 0.7 waste heat off the turbine can easily move the 0.3 units of data center electricity out of the data center and 0.4 units of waste heat (+ 0.3 data center heat) can still be used for another purpose. It might be good for a data center operator, but from a systems perspective the better use for that CH4 is still in a combined cycle utility plant which can make 0.6 electricity, use the waste heat for some co-located industrial facility and make the datacenter run an electric AC (COP ~ 3).

Re:

[Bob-taro]

I read the article and I'm still not sure I understand:

...and that heat is converted into chilled water using a liquid cooling system of absorption chillers that IBM and SU created.

So, is that saying the turbine heat powers something like a heat pump that then cools the computers?

Re:

[falzer]

See http://en.wikipedia.org/wiki/Absorption_refrigerator [wikipedia.org]

Re:

[sgt scrub]

hmmm that makes me wonder. Could the waist heat from a data center move up a chimney fast enough to move a turbine without the additional heat?

Re:

[thejaq]

Yes, but probably not worth it. See solar updraft tower.

Cooling with gas

[SEWilco]

heat output can paradoxically be harnessed to cool the servers

Someone's never heard of LP gas-powered refrigerators [wikipedia.org].

can do the same with kerosene

[Chirs]

I had one overseas where we had no power.

paradoxically?

[Anonymous Coward]

There's no "paradox" in using waste heat to generate chilled water. It's done all the time.

Thats what I do.

[sgt scrub]

I do this at home. My computers are really cool. Oh wait. Natural GAS not natural grass. Nevermind.

obligatory Morbo

[shipbrick]

Windmills do not work that way! Goodnight [youtube.com]

Perpetual motion

[tehcyder]

Once we've got that finally cracked then things will get really interesting. Who knows, we might even have true AI within twenty years.

Morbo

[Baloo Uriza]

Windmills do not work that way! Goodnight!!

One of these...

[Bartles]

Turbines can do a lot of work. They can produce power and compressed air simultaneously. Would be interesting if that compressed air supply was run through one of these. http://en.wikipedia.org/wiki/Vortex_tube [wikipedia.org] We have a small one in the shop, a very simple device. Connecting it to our shop air, at 125psi, creates air at one end that will make your fingers go numb, and the other end outputs heat at about hair drier temperatures. The tube is about 6 inches long and 1 inch in diameter.