
Free hot water?
Who pays for the electricity to power the attached server?
Heata has developed a novel way to use the waste heat generated by servers: mounting them on domestic hot water tanks to cut energy bills for homeowners. The company, whose services now include cloud compute, 3D rendering and ways to help landlords improve the efficiency of their housing stock, has hooked up with cloud …
According to their website:
"Depending on your supplier, we’ll either directly credit your energy account, or your bank account, for the cost of the electricity that the heata unit uses, and it will reduce the time your hot water heating is on. So a good portion of your hot water will be free during the trial. The exact amount is what the trial is trying to work out."
I'd imagine the server cooling is pretty inefficient when the tank is near an operating temperature of 55 to 65oC. (Don't run your cylinder cooler than 55oC unless you want to create an environment which might grow legionella bacteria, and personally speaking, I wouldn't run it below 60oC)
A small temperature gradient leads to very inefficient heat transfer and/or an overheating server. (Especially since the hot water cylinder will almost certainly be in an enclosed space. I'd be very interested to see the heat transfer calculations justifying the design.
Using a server as a wall mounted fan heater makes a lot more sense, and was tried in the Netherlands a few years ago, if I recall correctly.
I don't think this is about heating. The key question is: how do domestic electricity prices compare to commercial prices?
I'd have thought that big consumers of electricity (i.e. data centres) could get their electricity considerably cheaper - buying in bulk, direct access to wholesale markets etc. This would surely wipe out any benefit from "free cooling" which could be obtained by installing servers in people's homes - assuming it can even be made to work against the temperature gradient, as others have said(*).
However, right now, domestic electricity is cheaper due to government intervention in the marketplace. Is this what the companies are really trying to exploit?
Once this loophole is found, regulatory mechanisms will have to be added to forbid domestic consumers reselling their electricity.
(*) Not to mention the problems of managing and maintaining such a distributed server fleet, and the few workloads which are suitable for a server with poor network connectivity. And having to shut the server down when the water tank reaches its maximum temperature.
Hi,
The water comes into the tank cold and we don't touch the heating already in place so no change the legionella risk.
Ultimately British Gas are happy here and we've been running units for a few years, so we know they work.
P.S. Fan heaters make no sense in the summer, a seasonal cloud compute company would be fairly tricky to sustain :)
They probably worked this out, but how do the cool the server? Supposedly the element is still in place to heat the water in the geyser to 55°C / 65°C / 95°C, or whatever the landlord feels like, so won't they just be heating the server? And what good will it do to attach the server to the outside of the geyser? The one in my house has a thick layer of insulation between the tank and the cladding/hull/outside sheet metal thing.
The patent does not inspire confidence...
"The method may comprise pushing a frame through the insulation to cut an aperture in the insulation. A flange on the frame may define the depth to which the frame is inserted. The method may further comprise removing the insulation from the aperture and exposing a surface of the tank."
I hope they mean fiberglass blankets, but I imagine some installer will be keen to see how far he can shave down the side of the tank for a better conduction.
Not necessarily - most tanks are insulated with polyurethane foam (which is a nightmare to locally clean to give a good thermal contact with the copper).
Well, those that are left anyway, as stored hot water systems are ripped out for replacement with more efficient on-demand combi boilers ... then replaced as people have air source heat pumps put in so they can electrically heat their water only to find how much it costs for the electricity to top-up heat the inefficient stored hot water system! Doh!
Seems like more than a sniff of a snake oil product to offload the expense of the a data centre to me.
"Because it's a government-backed trial, we're putting the units for free in people's houses. And they'll be getting free luke warm water from it for a year" - FTFY
So this seems like a useful way of extracting money from the government while taking up space in people's airing cupboards and siting computers in very dusty environments.
I don't know about the economics or viability, but if cold water or cold return loop comes into the heating element at x degrees and need to be heated to 60 degrees then the heating system has to pump in energy to add 60-x degrees to the temperature. From my understanding of this project, the cold return will be pre-heated by y degrees so the primary heating system no only needs to heat the water by 60-x+y degrees and so has to put in less energy. If y costs the end user nothing then they save y energy costs.
As others have noted: the servers require electricity to run; electricity is generally considered to be the most expensive way to generate heat. Furthermore, boilers are on the way out as a way of heating buildings with heat pumps acting as inverse air conditioners to extract warmth from the environment. I suppose you could do that here but it would be the equivalent of driving one with a hair dryer.
But, as a business proposition, there is also a glaring risk over the data security and server reliability. Can't imagine any corporate lawyer or security team signing off on this.
No, far better to connect data centres to district heating (and perhaps cooling) systems and take advantage of scale.
But, given all the problems, does it surprise us that this gets UK government funding?
Would it not currently be cheaper to run these in a home since that would be on a domestic (capped) electricity tariff if they paid back the householder for the energy consumed compared to business tariffs which are currently much, much higher?
Not saying this concept is a good idea, just an interesting thought exercise.
As to dust which someone mentioned you'd only get dust with active cooling via fans, I thought the whole point of this was that it was a sealed black box so all the heat could be transferred to the tank?
Pretty sure I could arrange this myself if I really wanted to, there are services to rent out CPU , GPU time, etc.
But I'm not sure I want to pay electricity to run an inefficient heater to generate lukewarm water from a server based on a water tank in my loft that has to be somehow attached.
Even a loft itself is a poor environment for a computer - dusty, spiders, condensation, extremes of cold and heat, inaccessible, potentially rodents etc.
I bought a house recently with loftspace that I don't use for storage (I don't have enough stuff to justify it), but I still sited my "servers" (Raspberry Pi's), NAS and network cabinet in a cupboard rather than in the loft. Also, I chose a cupboard that's NOT under the water tank. I've never had a leak, but I'm not taking that chance.
In actual fact, some of the cabling (to cameras, network sockets etc.) does go into the loft - none of it near existing cabling or the plumbing deliberately.
The "waste heat" from that cupboard just stays in it. It's not enough to worry about and it heats the house as a byproduct, and heating air is a lot easier for a computer than heating water. They even come with fans built-in...
The days of random people running a computer unnecessarily to sell the compute time are long gone, several cycles over. The last was the rush for SETI/Folding@Home, etc. It's lovely that people want to donate their time and money to projects like that, but you'd be better off dropping some cash into an AWS server instance or even just renting a cheap dedicated server or similar, in terms of value for computing.
Getting enough money to pay the homeowner for the necessary several hundreds of watts to slightly warm their water? I can't see it happening, let alone covering, installation, management, decommissioning, etc.
You'd be better off just buying a solar panel and attaching it to a 100W heating element. Sure, it wouldn't get 100W all the time, but it would do more directly and cut out several middlemen. And the cost of a 100W or even 200W panel is going to be less than whatever computer you're installing there.
Hell I have heating that runs off a homebrew solar system, I could do a better job just plugging in a small, low-power water heater (e.g. a fish tank heater dangled into my header tank) than I ever could trying to do this with the computer middle-man.
I totally understand where you are coming from, but using a solar panel is a new installation generating new power. The proposed heating system is using heat that is being generated anyway and would normally be exhausted to the outside world and wasted. Whether it's economic to distribute the heat in this way rather than aggregate it in a data centre and use the waste heat at economies of scale is the real question.
This kind of works but also doesn't when you get into the weeds.
I can get that Qarnot have a long term business case. Their 4kW units are often doubled up at commercial sites and concentrate a lot of compute sleds together. Almost like a mini data centre. They're going to be easy to renegotiate access to and much, much quicker to update hardware or maintain.
Heata have more of a one shot strategy. Those units are going to come to the end of their life in terms of the compute you can sell in a few years, maybe seven at a stretch. Then there's thousands of site visits to sort out with not a huge amount of financial benefit for each visit. Hopefully just the board & SoC inside changes not the whole chassis and heat transfer adapter. New owners house owners being confused could be an issue too.
I'm all for this kind of heat recovery, I just think that GleSYS community heat supply and Qarnot's commercial building heaters make more sense.
The potential cost savings doesn't come from electricity arbitrage, it's from not having to build a data centre, Real estate is expensive to acquire, develop and maintain. If you rent, you are paying someone else for all those costs plus whatever they need to make it profitable.
The question is whether install and service (aka man with van) of your distributed data centre is cheaper than those reat estate costs. Which I imagine is what this trial is trying to figure out.
How efficient is this thing going to be?
Obviously the "heating element" will be 100% efficient, like any electric heater, but how much of that heat is going to make its way into the water? If you strap a heating element to the side of the boiler, my guess is, very little, which is why the heating elements for electric water heaters tend to be inside the water rather than strapped to the outside of an insulated container.
If it was a water-cooled system with the hot water tank as the pump reservoir, then that might be more efficient, but would you want to run regular tap water through it without any additives?
> how much of that heat is going to make its way into the water? If you strap a heating element to the side of the boiler, my guess is, very little, which is why the heating elements for electric water heaters tend to be inside the water
The heat has to go somewhere. Dense water or thin air? With any decent contact to the water tank (curvature a problem) IMHO a vast majority of the heat will go to the water. (And thus defer the need for dumb-electric heat.)
Dumb-electric heat elements are made as small as possible because cost. You can't put kilowatts in a hand-size space unless it is directly liquid cooled. (Or steam, but I gather typical electric elements are not made for the thermal shock of boiling.) In other fields it is quite common to make "electric blankets" which wrap around and strap to a tank. Less intensity of heat (but more area), cheaper construction, no large bung to rust and jam, possibly more shock hazard through thin blanket..... swings & roundabouts.
I think it is daft, but not because of thermals. (Do I really want BOFH or PFY or their spawn coming into my cellar or attic at odd hours to reboot a stuck server? Will my loud-mouth dog object? Do they bring beer?)
It's nice to see the company represented here in the comments - At least, from what he is saying, I assume it is the same Chris Jordan.
Chris, how many units do you have installed so far and how is it tracking compared to projections?
Hi,
Thanks for the question.
We currently have 20 units in the wild, soon to be nearly 100; 80 as part of this trial.
How much makes it to the tank is depends on a few factors, but the units currently in houses put between 180W and 200W into the tank when at full power. For ref. these units are twin 120W Xeon CPUs.
200W * 24hours give 4.8kWh - This is just above an average household's HW usage. Then, allowing for down time and other factors we hope to supply ~80% of a household's hot water. Not all, we don't replace the current system; we're like solar thermal in that respect.
This was already known from a previous trial, but we needed a larger scale trial to provide stronger evidence for an application to be part of the system for calculating EPC ratings, they are rightly very cautious when it comes to product claims.
thanks, I hope that helps
I presume this requires enormous tanks. I have a heat pump hot water heater that consumes about 500W of electricity. That's over 2000W of heating in summer when the garage is a solar cooker but maybe only 600W to 1000W in winter. That low level of recovery means double-sizing the tank to minimize resistive heating.