Data furnaces arrive in Europe: Free heating, if you have fibre Internet | Hacker News Reader

Data furnaces arrive in Europe: Free heating, if you have fibre Internet

arstechnica.co.uk

120 pointsriffraff72 comments

There's a datacenter in Helsinki that pushes its heat to a district heating system.

http://www.theguardian.com/environment/2010/jul/20/helsinki-...

Nice article. It certainly works at scale because of the economies. I think this distributed system has potential but there are some issues. There are also some things they could do to make it better if the trial is a success.

Using resistive heating (which is what this is) is not very efficient (even if it's a by-product). A better way to heat your home is to not waste power and use it instead to run an electric heat pump (which are over 100% efficient). There is a good analysis in the "Mythconceptions" section at the end of chapter I.11 of the excellent "Sustainable Energy — without the hot air" [0] on p71 (p84 of the free PDF).

That's the downside but there are some interesting applications. They could enable you to host your own personal cloud on the box in your home. So you know where it is.

You could vary the operating time of the servers so they only operate (or at least only run at 100%) when heating is needed. This could be based on the season, an external temperature sensor or weather data. Processing could migrate between the hemispheres of the world depending on the season.

The load could also be varied based on the local grid demands and power mix. Various real-time data sources are available that report on how green the current energy mix is [1] [2]. You can even just measure the frequency of the electricity to see how well generation is matching consumption. This is known as dynamic demand or demand shaping and you can even be paid by the grid operator for it so it could be a good business model. It would also be nice if they partnered with green power companies that buy their energy from renewable sources.

[0] http://www.withouthotair.com/download.html [1] http://www.earth.org.uk/_gridCarbonIntensityGB.html [2] http://www.gridwatch.templar.co.uk/

> Using resistive heating (which is what this is) is not very efficient (even if it's a by-product).

If heat is a byproduct it is highly efficient to actually use it rather than waste it. Here the electricity would be used either way for the processing tasks. The system would be inefficient if it just ran busy wait to produce heat, but that's not the case.

> A better way to heat your home is to not waste power and use it instead to run an electric heat pump (which are over 100% efficient).

"100% efficient" makes no sense, let alone "over 100% efficient", you're talking about coefficient of performance (the device's energetic ROI). If you ignore what the system is actually doing, the eRadiator thing has infinite efficiency since you're not paying for any heating (aside from installation costs, which you also have to pay for a heat pump)

The over 100% efficient means that the heat produced in the room is greater that the energy taken from the wall plug. This makes sense for me.

That's not how thermal efficiency works. And if you don't care for actual thermal efficiency, the energy taken from the wall plug is irrelevant if you're not paying for it.

http://en.wikipedia.org/wiki/Thermal_efficiency#Heat_pumps_a...

The expression "coefficient of performance" is used instead of efficiency for heat pump.

There's also one in Stockholm, Sweden owned by Bahnhof AB: http://www.opendistrictheating.com/pilots/bahnhof-thule/ (From the same company that also has the "James Bond Data Center").

...And no data cap, and don't mind a chunk of your bandwidth being used, and don't mind a locked box with an always-on and encrypted connection to who-knows-where in your house. I'd also worry about what happens if/when it breaks down. (For instance, if the power supply fails. Or whatever: there are pretty much always single points of failure on such things.)

A neat idea in some ways, but I'd be wary.

> and don't mind a locked box with an always-on and encrypted connection to who-knows-where in your house

Do you fear hidden microphones, cameras and other sensors? If not, then what's the harm?

Since you have no control over it (aside from ripping it out of the wall), you're at the mercy of whoever runs the box to adequately secure it. When it gets hacked, any shady business conducted by the miscreants will be emitted from /your/ IP, so you're exposing yourself to TOR exit node level liability.

You might be able to work around that.... but it depends what's running on the box. If it's a backend server for some organisation (database, etc) then you could have some sort of tunnelling router between it and the fibre that routes all packets back to the organisation's datacentre.

I worry about the potential consequences of legitimate sensors being repurposed, yes.

Network snooping / spoofing / etc (even if it's on a VPN through an external router it still leaks congestion information!), temperature settings, especially versus actual temperature, power fluctuations, vibration sensors, etc, etc.

I don't particularly want random people knowing - and potentially selling - information about when I sleep, when I eat, when I take a shower, when I'm not home, when I'm on vacation, etc, etc.

And don't mind someone else accessing the internet through your account. Better hope they aren't running a search engine that stumbles on kiddie porn/extremist material/pirated content, etc...

As long as it's set up so it can only connect to the master thing, preferably on a shielded and separate internet connection, then it should be fine - legally, it's their device.

They'd probably have a few service employees doing regular maintenance and replacements on the road though. That would probably be the biggest expense they'd have after installation, actually.

Not just legally.

Another question is what happens if/when they get hacked. Remember: to you, it's a locked box. You have no control over their security - or, for that matter, for the security of whoever may buy the company in the future. Much less them deciding to sell your data sometime in the future. (Remember RadioShack?)

Why would the security matter to you? It's the owners problem...

Because there's an astounding amount of data leakage from something like this.

For instance:

I don't particularly want random people knowing - and potentially selling - information about when I sleep, when I eat, when I take a shower, when I'm not home, when I'm on vacation, etc, etc.

See also "Heating houses with 'nerd power'" from last week at https://news.ycombinator.com/item?id=9581588 .

What about the electricity costs to run the machines?

It's a cost-sharing agreement, so electricity costs are split between the data-furnace provider and the home-owner.

According to the efficiency of the device?

The heating efficiency has to be 100%, because the electricity isn't being converted into any other energy form.

(That's still worse than an electric heat pump though, though - those have > 100% heating efficiency in terms of electricity input).

Everything above 0K will emit photons, no?

Right, but at those temperature the photons have near infrared frequency, hence they won't escape your house and simply heat it

There will also be plenty of RF interference, although that's at a lower energy, and therefore will use less energy to produce.

> (That's still worse than an electric heat pump though, though - those have > 100% heating efficiency in terms of electricity input).

"Coefficient of Performance" being the over-100%-not-technically-efficiency measure for heat pumps. [0]

Or "Seasonal Energy Efficiency Ratio" if for god knows what reason you want to measure it in British thermal units per Watt hour. [1]

So a setup like this isn't actually competitive with a heat pump in most conditions, but if the electricity was going to be burned in a datacenter anyway you might as well use the heat for something productive.

https://en.wikipedia.org/wiki/Coefficient_of_performance

https://en.wikipedia.org/wiki/Seasonal_energy_efficiency_rat...

If I understood it correctly, there is a server running inside that thing. So by definition, all the heat being generated is inefficiency. Therefore, if 1000 W are being disposed as heat to the environment, then the machine itself is consuming WAY more than that. Otherwise, they found a way to do computing without consuming any energy.

edit: Upon further googling, it looks like I was stupid here. So almost all energy consumed by a PC is dissipated as heat? Did not know that, sorry.

So almost all energy consumed by a PC is dissipated as heat?

Yes. The other forms of energy leaving your PC range from tiny to infintesimal: light from the keyboard/case LEDs, kinetic energy in the exhaust air molecules, radiated RF (both intentional like WiFi and unintentional), radiated sound (both intentional from inbuilt speakers or connected headphones and unintentional from vibrating components), earth leakage, and resistance losses in external wired network connections.

Most of these forms are then reabsorbed and converted to heat by other objects in the immediate surroundings as well.

Increases in the energy stored within the PC are entirely dominated by temperature increases (which will equalise with the surroundings over time). There would also be a truly infintesimal amount of energy stored in changed potential energy in magnetic domains on your hard disk or stored charges in your SSD.

Yes, I get it. Being a Mechanical Engineer and used to balancing energy in heat engines, I just (incorrectly) assumed that SOME kind of work was being done by the computer to perform its main function, which is obviously not the case (at least not to non-negligible amounts). Thanks for clarifying.

Ahh, as an ME you would appreciate then that the useful work a computer does is not thermodynamic work but a decrease in the disorder of some system it is a part of, and this necessarily requires increasing the entropy of its environment - typically by transferring heat to it!

Some energy (but not much!) goes out through the network.

Half the time it would be "off" (during summer), so that means I will pay for 100% of the actually used energy? In that case I'll stick with natural gas for my heating. Also this arrangement should not be called "free".

It's installed on an external wall so when you turn the heater 'off', the heat is redirected to outside. They will apparently also reimburse you for energy used.

That's what I mean, off means it's still "burning" but I get no benefit. And according to ihsw the bill is split leaving me with: In winter: Gaining 50% of the power in heat (don't know how that compares heat-wise to natural gas) and in summer, wasting 50% of my own paid-for electricity. I actually doubt if ihsw's comment is true.

The bill is not split -- it says they reimburse you for the measured power usage of the device. Nowhere does it say that this only applies when the heat is being directed into your house, that I could see.

I don't think that's right, their FAQ says

"Can I really heat my home for free using your heater?"

"Amazingly, yes! The Heater measures how much energy it uses to heat your home and you’ll get fully reimbursed for the cost of this energy. This means that with our Heater you’re heating your home for free!"

http://www.nerdalize.com/faq/

I can't find anything about a cost sharing agreement, the only cost I can see is the upfront purchase price in the BBC article (although nothing on the nerdalize site itself).

Can you turn them off? Obviously you don't want to run a heater in the summer.

And if so, wouldn't the resulting (massive) underutilization of hardware erase whatever savings the data center would realize from not having to deal with heat?

Not to mention the other (massive) costs of remote servicing, like literally rolling trucks to replace a hard drive. And 1000W is not even one standard space heater. I like the general idea but this decentralized in-the-home concept seems about as well thought out as their logo.

It would be better suited to a centralized application like the heated swimming pool: year round usage, enough density to achieve economies of scale in servicing and providing site security.

Or how about this: locate your next data center near a town with a district heating system, so you can tap into it and pipe waste heat directly into homes. (Sweden has lots of waste-to-energy plants that burn trash to generate electricity and hot water heating distributed through networks of pipes.[1]) In the case of Sweden you wouldn't even have to build out much infrastructure; it's already there. I wonder if anyone has explored this?

EDIT: I just saw ville posted about this very idea. Cool.

EDIT2: jessaustin notes that they keep the server running constantly and probably use a powered heat pump to vent it outside during the summer. (Which still raises questions for me about the efficiencies.. small scale heat pumps probably aren't as efficient as a large-scale data center implementation.)

[1] http://en.wikipedia.org/wiki/District_heating

The article mentions that the furnace vents waste heat outside when you don't want it. So, yes, it's running all the time.

> Or a fuse blows?

Or a kid pours his cup of sticky soft drink in the vents (heaters need to have vents right, almost certainly at the top). Anyone who thinks something can be "tamper proof" has never lived with a 4 year old.

They surely use redundant, encrypted storage. MaidSafe is a distributed "cloud" storage protocol that does that: http://maidsafe.net/

Cloud&Heat are another company doing this in Germany: https://www.cloudandheat.com/en/index.html

And a French one: http://www.qarnot-computing.com/technology Qarnot computing

A pun on on Nicholas Carnot, the creator of this: https://en.wikipedia.org/wiki/Carnot_cycle

Another French one: http://www.defab.fr/

They're heating water for consumer purposes. I met one of their founders yesterday, they're pretty neat!

It would be very interesting to hear their security analysis and conclusions. The article's casual use of the term "tamper-proof" doesn't sound credible without details.

The reliability issues that are mentioned are no problem with enough redundancies of course.

My fiber optic connection is 50/50 even though the line is (the way it is used now) able to do 500/500. They should aim for deals with ISPs for my excess bandwidth.

When it's in the "push heat outside" mode on a hot summer day, there is a significant amount of air to pump, in order to dissipate a kilowatt. I wonder how noisy is the unit in this case.

I heated my room in college for 4 years using nothing but a Q6600, never turned on the radiator once. Nice solution, but probably will have issues cooling for anywhere that has 90F summer temps.

How about the noise?

Well that is a different story because I was using 4800rpm 80mm fans like an idiot back then.

I'm doing basically the same thing at the moment with a rack of xServes and an HP Proliant. Pretty quiet actually, compared to my music and other noise in my area.

It must be pretty warm where you live. I had a couple of Q6600 upped from 2.4GHz to 3.2GHz, and that wasn’t enough to heat the room. It was in Russia though...

> It must be pretty warm where you live.

Good insulation and small surface also helps. You won't heat up a 2-story house with just a computer, but a well-insulated dorm room is more than feasible.

I always thought of Bitcoin as an implementation of this idea.

What happens in summer?

It still heats of course. It's free heat.

In the summer, in most locations, you don't want free heat. However, the article states that the unit will vent the heat outside when it isn't desired.

I've heard of summer.

  if you "turn off" the eRadiator, the servers don't actually turn off; the heat is just pushed outside

Same as in a data center, the heat gets pushed outside.

TFA:

To be eligible for the eRadiator, your home has to have... "an external wall" ... and the external wall is needed for venting (if you "turn off" the eRadiator, the servers don't actually turn off; the heat is just pushed outside).

Admittedly, without active refrigeration this is unlikely to work too well when the temperature outside is over 35 C.

EDIT: they don't mention separate electric supplies, so presumably the unit has a way of tracking its power usage and reimbursing the homeowner. Otherwise this is just a scam, because far more efficient electrical [EDIT: thanks 'votingprawn!] heat pumps are available for less initial cost. However, the non-scam scenario seems only to make sense as a business in cold locations.

Makes complete sense in Scandinavia, Iceland, Alaska, etc.

Is this a correction? A clarification? Was it insensitive to describe those places as "cold"?

A clarification. For instance, there's plenty of cold in Antarctica, but no fiber and not enought dwellings. In Scandinavia, high fixed broadband penetration is actually combined with cold most of the year. (JFYI, Finnish official definition of summer is temperatures above +10°C for entire 24 hours.)

> because far more efficient electrical heaters

How are you defining efficiency here?

Efficiency is usually defined as (useful power out)/(power in), and basically all the power that goes into a computer comes out again as heat, which in this case is considered useful. Not really convinced you could buy a heater that was measurably more efficient.

ETA: heat pump would do it, didn't spring to mind when I read heater though!

> seems only to make sense as a business in cold locations.

I'd change that to read "in cold or temperate locations."

There are plenty of locations in Europe for example that rarely have days over 30 C yet don't often get below freezing in the winter. Doesn't make sense to describe these places as particularly "cold."

Cloud computing users don't care---modulo network latency and, to some extent, national laws regarding data privacy and seizure of servers---where their servers are physically located. When it's summer in the northern hemisphere, it's winter in the southern; I'm sure they're looking at South America too.

Sure, but I don't think the provider can afford to buy servers and only run them for half the year.

International speeds on a consumer grade line are usually crap, around 10% of declared speed.

This could be a supplement for existing systems, perhaps even a backup or to serve some use spikes. But definitely not a replacement.

This is best suited for async workloads (think EC2 spot instances or cpu intensive distributed computing tasks) where network performance isn't critical.

We just need to roll our own clusters of RPi-s driven off Peltier junctions arctic-silvered to the side, and we can run our own parasitic cloud provider off the waste heat. It's turtles all the way down...

I assume next versions will have AC?

There are only few days a year when it gets over 30 in Scandinavia & Baltics.

Probably more like a few days a decade when it even approaches that here in Scotland!

This seems ideal for running TOR exit nodes. Except for the home owners of course. I wonder if their TOS forbids this?

Reply on news.ycombinator.com