Weekly Edition
Return to the Press page
| |||||||||||||
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal)
James Gray is
looking for creative ways to recycle waste server heat. "The
botanical garden that is heated by Notre Dame's servers specializes in
desert plants and is located in Indiana, well know for its brutal
winters. The concept is simple. The waste server heat is pumped into the
interior space of the garden to keep the desert-loving plants toasty warm
all year long. This simple, creative step is saving the university $100,000
on cooling costs and the owner of the botanical garden, the City of South
Bend, Indiana, another $70,000 on heating costs. Not only that, but the
atmosphere is spared many thousands of tons of pollution from carbon
emissions. That is quite a triple win-win-win arrangement."
(Log in to post comments)
Thermodynamics says it can't be a win-win Posted Oct 14, 2008 18:35 UTC (Tue) by martinfick (subscriber, #4455) [Link]
Someone is not being upfront about the costs. Pumping your extra heat into the botanical gardens is very unlikely to be actually saving money on cooling. There is probably something else that they have done to save on the cooling if they are indeed saving. But simply pumping the heat into the cold Indiana outdoors would certainly save them more money on cooling then pumping it into the gardens. So, to be honest, the cooling savings should be negative, but it can still be valuable if the loss from the cooling bill is outweighed by the savings on the heating bill. But this cannot be a win-win situation, someone is obviously piling it on high and deeper.
Thermodynamics says it can't be a win-win Posted Oct 14, 2008 19:09 UTC (Tue) by xbobx (subscriber, #51363) [Link]
If the garden used to pay $170k for its heating, but now pays Notre Dame $100k for its heat that would otherwise be vented outside, then Notre Dame can now put that $100k towards its cooling thus "saving" $100k on Notre Dame's cooling bill and $70k on the garden's heating bill.
A bit oversimplified I'm sure, but it's certainly possible for such a mutually beneficial arrangement to exist.
Thermodynamics says it can't be a win-win Posted Oct 14, 2008 19:27 UTC (Tue) by martinfick (subscriber, #4455) [Link]
Well, that's not a savings on cooling, it's simply gaining a revenue source out of the excess heat. The actual cooling did not get more efficient, in fact it probably took a hit. In other words their cooling energy bill actually probably went up, but it is now being offset by the extra revenue!
As you pointed out, economically it can be a win-win for both partners, just not thermodynamically. The bottom line is good.
Thermodynamics says it can't be a win-win Posted Oct 15, 2008 20:01 UTC (Wed) by jordanb (guest, #45668) [Link]
Well, in the whole system energy has been saved, because previously the heat from the computers was being vented as waste, and then more heat was being created (by burning natural gas or heating oil or what have you) by the botanical garden to keep its temperature up.
Thus the whole system *is* more efficient. Less heat is being vented as 'waste' and less energy is being consumed by the system containing the data center and the botanical garden.
Suppose your car had a heating coil to run the cabin heat, and then someone said "let's use waste engine heat to heat the cabin." Well, now the car will get better gas mileage because it's no longer running the engine *and* the heating coil, so that's an efficiency gain.
Thermodynamics says it can't be a win-win Posted Oct 14, 2008 21:35 UTC (Tue) by dkite (guest, #4577) [Link]
I'm imagining such a system which would have a water loop where in the
winter the heat is put into the greenhouse, and in the summer is diverted to an evaporative condenser type of situation. You can gain efficiencies in cooling operation in the summer with this type of system. You are rejecting heat not only by sensible means, but latent as the water evaporates, giving you a lower medium temperature than you could get with a straight sensible heat exchange.
Even winter operation could be better due to the way head pressures are
Derek
Thermodynamics says it can't be a win-win Posted Oct 14, 2008 21:53 UTC (Tue) by martinfick (subscriber, #4455) [Link]
Your evaporative cooling system is good, but not related to what they are doing. And the summer is when they would actually pay less of a penalty to pump their heat into the gardens instead of the outdoors. Of course, in the summer, they are less likely to get paid as much for the heat since the gardens will need it less.
Any minor gains due to constant loads would be extremely offset by the huge temperature differences of pumping heat into a cold winter outdoors (efficient, no compression even required, simply circulation) versus a hot desert like indoors (very inefficient)!
Thermodynamics says it can't be a win-win Posted Oct 15, 2008 0:27 UTC (Wed) by dkite (guest, #4577) [Link]
As for winter looking after cooling, it's not that simple. The energy
produced in the large space needs to be transferred somewhere. The system to transfer the energy has to work in the winter, shoulder seasons and summer.
Typically the DX systems used are prepared for winter operation by putting
The industry has changed quite a bit over the last 5 years, and gotten out
Which in the end would save on cooling costs in summer and winter.
Derek
Thermodynamics says it can't be a win-win Posted Oct 15, 2008 0:32 UTC (Wed) by clugstj (subscriber, #4020) [Link]
Evaporative cooling is only worthwhile in dry climates, which Indiana certainly is not.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 14, 2008 20:59 UTC (Tue) by BrucePerens (guest, #2510) [Link]
I think mine heats my office. I taped over the heating duct, and it's still warm enough there in the Northern California winter.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 15, 2008 2:38 UTC (Wed) by dlang (✭ supporter ✭, #313) [Link]
California Winter
what do these words have to do with each other ;-)
I know, you specified Northern California
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 15, 2008 4:17 UTC (Wed) by flewellyn (subscriber, #5047) [Link]
Try it out in a North Dakota winter, and then we'll talk.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 15, 2008 5:09 UTC (Wed) by BrucePerens (guest, #2510) [Link] I teach in Norway, I'll be there in December. Not much sun there all winter.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 16, 2008 10:19 UTC (Thu) by ekj (guest, #1524) [Link]
But much of the coast nevertheless has a mild climate, here in Stavanger on the west-coast for example, normal winther-temperature is around freezing, which ain't horribly cold, methinks.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 15, 2008 7:51 UTC (Wed) by PRB (guest, #54702) [Link]
I'm always thankful for journalists sharing new ideas in sustainable/energy efficient design. Here is the technical report on the grid heating project, I'm the fellow in the picture. To clarify some of the press releases this project has completed phase 1 (demonstration of capability) and is moving toward phase 2 (scaling up the number of servers to meet larger heating requirement). ND's HPC utility budget has rapidly grown from 50K toward 100K, the greenhouse heating bill is over 100K. Seasonal applications have obvious limitations in efficiency, we are actively collaborating on year round hot water production prototypes. The focus is definitely reutilization of the heat. Seasonal relief/make-up air utilizes well documented free cooling benefits. Year round applications focus on heat pump and similar technologies.
http://www.cse.nd.edu/Reports/2008/TR-2008-09.pdf
Regards,
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 15, 2008 8:55 UTC (Wed) by wouter (subscriber, #10160) [Link]
This is just so wrong; why generate all that heat in the first place? We should really focus on minimizing energy consumption instead finding arcane applications for waste heat.
Does not work this way Posted Oct 15, 2008 18:04 UTC (Wed) by khim (subscriber, #9252) [Link] If we'll manage to produce 2N computation for the same X joules people will build the facility to produce 4N computations and we'll be even worse off. Devon's paradox, you know... The only way to really solve this problem is to raise cost of energy - it'll happen over time with or without our help...
Does not work this way Posted Oct 18, 2008 0:14 UTC (Sat) by giraffedata (subscriber, #1954) [Link] I agree completely that simply making computers more energy efficient doesn't mean we'll use that much less energy in computers.
The only way to really solve this problem is to raise the cost of energy You didn't say which problem, but I assume you're talking at least in part about global warming. While I agree that the only good solution involves higher energy costs to users of energy, that should be a result, not something we do. What we should do, for greatest efficiency, is attack the problem closer to its source: people pumping carbon dioxide into the air, where other people suffer the cost (i.e. the global warming). People who pump carbon dioxide into the air should pay a tax to cover the atmospheric clearness they're using up. That would result in higher energy prices, which would result in more efficient computers and less computing.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 15, 2008 18:38 UTC (Wed) by tialaramex (subscriber, #21167) [Link]
Minimising energy consumption is the cheap, easy option. So you should not be surprised that most people have already tried that. Once you've exhausted the cheap, easy options, you have to get a bit more creative. This is certainly creative.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 16, 2008 15:16 UTC (Thu) by NRArnot (subscriber, #3033) [Link]
Here in the UK, buildings are heated during the winter months, but I have yet to hear of a server farm being coupled into the containing office's central heating, so that no heat it generates is wasted for the half of the year that the office needs heating. It would probably work better in the Northern USA where the outside is a lot colder. Why aren't all Google's server facilities located in Alaska or near the Canadian border?
The cost would be a more expensive air-conditioning plant. It would need two heat pumps, a conventional one coupled to the air outside, and a new one pumping heat into the building's heating system. This second one would be less efficient at moving heat, but since all the pumping energy would also end up in the hot water, in another sense it would be 100% efficient.
If energy gets expensive enough, and datacomms and CPUs cheap enough, one could even have a northern hemisphere data-centre and a southern hemisphere data-centre, and operate them only during their respective cold seasons!
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 16, 2008 20:02 UTC (Thu) by dmills (guest, #54200) [Link]
How about using the waste heat to preheat feed water for a local office or hotel buildings domestic hot water systems?
This has the advantage of good thermodynamic efficiency as you are probably not pumping heat up a temperature gradient (the buildings main heating system can take the water the rest of the way).
It seems to me that the art here is to find a use for large quantities of very low grade heat, rather then trying to produce high grade heat from the system.
In fact, if you think about a standard server, there are a fairly small number of components that generate most of the heat (in some cases potentially of quite high grade), would it be worthwhile designing servers to separate the high grade heat from the low grade stuff?
Power switching mosfets have typical maximum junction temperatures of 125 degrees C, but probably mostly run at far lower then that, and typically dump the heat into the general cooling air at 40 degrees or so. Similarly, the CPU chip is often good to at least 60 degrees C case temp (and in the case of something built using a modern mobile part, good to over 100 degrees C), having two cooling exhausts (one for the general low grade heat and one at 20 degrees C or so hotter for those parts that can handle it would allow the higher grade heat to be used efficiently for some applications that the lower grade heat just would not do for (Worst case, it is easier to get rid of).
At least part of the problem is to design the servers to be able to dump the heat as efficiently as possible and thus to be able to be run the ambient temperatures as hot as possible, having ambient limited by a 30 degree delta T across the CPU heatsink (with another ten across the chip/heatsink interface) is not good.
I could even see a server design with a conduction cooling mating surface for the high grade heat, and racks with liquid cooled mating surfaces to allow the parts that can to run at 60 degrees C or above with the heat being dumped into some useful working fluid. I am unsure if 60 degrees would be sufficient to allow this heat to power an absorption chiller to cool the more sensitive components.....
In any event almost all the electrical power you put into that facility will have to come out as heat (The difference is the energy needed to change the amount of order on the disk platters, as close to zero as makes no difference).
Just some thoughts.
Regards, Dan.
Ideas Wanted: Creative Heat Recycling from Servers (Linux Journal) Posted Oct 17, 2008 23:59 UTC (Fri) by giraffedata (subscriber, #1954) [Link] why generate all that heat in the first place? Because botanical gardens, among other things, need heat. Would you rather spend extra money to minimize energy consumption in the server room (and rest assured it costs extra money, or we'd have done it already) and then have the garden spend extra money to generate heat for the plants?
|
|||||||||||||