Heating houses with high performance servers
nerdalize.comI'm actually inclined to take this fairly seriously.
I personally run a 36U rack of servers in my basement for learning and hosting for my business. It gets toasty enough that the basement temperature hovers around 30-35C.
On cold days we open the door to the basement and the warm air rises into the ground floor, giving a bit of warmth.
On warm days I have a temperature controlled fan which vents the air to the outside.
My house has central heating but if it had central air I'm pretty sure an exhaust from the top of the rack to the central air unit would be do-able (not sure what HVAC building codes I'd be compromising though).
Puget Systems did a great blog post titled 'Space Heater vs Gaming PC'. They basically found they were tied in the efficiency of heat output from both devices.
Source: https://www.pugetsystems.com/labs/articles/Gaming-PC-vs-Spac...
It's interesting. There are obvious security/integrity problems, but there is probably plenty of not very sensitive data to process, and regarding tampered results, if it's cheap enough, one can always process everything twice and compare them.
That said, the eRadiator market seems to be limited to households in places which are cold all year and which have a decent Internet connection. What does that leave? Northern Europe?
We have summer two months per year thank you very much!
Their FAQ says it can still work when you don't need the heat in the house:
> Will the Nerdalize Heater work during the summer?
> Yes. Our Heater can expel excess heat to the outside when the homeowner does not require heating. This way we can compute at full capacity during winter and summer utilizing our hardware optimally.
That suggests, however, that installation is more complicated than just mounting it on a wall like you would a shelf or a picture frame and plugging in power and network. It sounds like you'd need to have something going through the wall to the outside.
They're essentially using your cooling in the summer unless it does air intake from the outside too.
This is a joke right?
Unix philosophy has long been that the person with physical access to the machine owns it. This would only be acceptable for a very specific class of compute jobs where nobody cares if you hack them. Like computation for charitable purposes. You could never host peoples personal or business data in this. Furthermore bandwidth would be limited and unreliable, as would power.
Not unix, physics.
Physical access means you can directly measure things that would otherwise be secret. Nothing stops you from reading the contents of memory wholesale, for one simple example.
A fully homomorphic encryption system would work, but I don't think we're there yet.
For some things we are, but I don't know how far along open source software is currently. It would take a major incentive to get people to even talk openly about what's possible in this space.
And produce a lot more heat, for the same computation!
CPU Technology like Intel Software Guard Extension or AMDs Secure Memory Encryption closes that gap as well. It's not magic and might be flawed in its current forms - I'm no expert on that - but I do believe in a not too far away future physical access will no longer provide cheap attack vectors.
Why ? They could embed some sort of a trusted platform module with crypto and (physical) intrusion detection, just like payment terminals.
If you open the case, the system deletes the keys and you can say bye-bye to your contract.
Physical intrusion detection can be defeated. All it takes is one person to open it up to document where the microswitches are, and a careful bit of work with some tin snips and you're in.
Yes all physical security can be defeated but you could make it exceptionally difficult to do. As an example, I used to work on financial terminals and sat across the isle from one of the security engineers. They actually embedded conductors in the case that created a known EM field when the case was closed. If you opened the case, it disturbed the field and erases the keys. If you drill a hole in the case, it disturbs the field. If you melt the case and insert a metallic object (screw driver or knife) it disturbs the field. This was just one of the physical security measures.
tl;dr - modern physical security is much more than microswitches.
Do these systems still hold up when an attacker has practically unlimited physical access to the device and you do not?
This seems like a very challenging environment for tamper detection.
I'd like to see that stand up to mom vacuuming right next to the unit.
I'm not sure. Say you have a few TBs of random sensor data to crunch, and you allocate chunks of a few GBs to each node. What can anyone in practice do with that? Especially if you have checks to avoid tampering with results (e.g. processing every chunk twice).
"Especially if you have checks to avoid tampering with results (e.g. processing every chunk twice)."
See, that's the problem every time this comes up (and it's come up about as often as micropayments). By the time you get done accounting for the unreliability, hostile end-users, lack of access, customer support, low-grade obsolete inefficient un-fixable infrastructure, increased avenues for attack, duplicate processing, super-high latency, super-low bandwidth etc, you've created something that's several times slower, more expensive, more unreliable, more complex, and harder to upgrade than if you'd just stuck it all in a datacenter where you don't have to propose insane mechanisms like double your computations.
Oddly, this isn't the first startup proposing this idea, french startup Qarnot Computing is also in this space: http://www.qarnot-computing.com
I'm still critical of it, but it'd be interesting to see if it flies in a constrained scenario of condos/hotels/office buildings - there's already scenarios where data centers will take up entire floors or basements of buildings and channel heat into the building's hvac system.
In theory, any heat generated without computation is a waste. I'm sure it's pretty tough to actually make that work at scale, so excited to see this!
Nice to hear, it is tough but we believe we have figured out how to do it effectively. Took a while and quite some prototypes/concepts to make it work.
A distributed version of this will take place in Denmark, once Apples new datacenter is online. We have a distributed heating network in denmark, where excess heat is "sent" to homes connected with "fjernvarme" translated "remote heat". This new datacenter will dump its heat into the existing heating network and will heat x homes in the neighborhod.
I think this is a much more sensible and useful implementation, where the computation happens in a centralized location but the heat energy gets distributed. The linked solution requires CPUs to be distributed for the the heat energy to go with them. As many other commenters post, this creates many questions about data security.
That system is not distributed, the system is centralized and the heat is distributed. We are a completely distributed system. This brings numerous economical advantages in terms of total investment required to roll out or scale up. We believe our way is superior to a heat network. Nonetheless the fact that the heat is reused can only be applauded.
New York has had a consolidated heating system for over a century
> fjernvarme
We Germans have that, too - Fernwärme :)
My 7x7 foot room at university was toasty with the combination of a GeForce 3 and Diablo 2.
I get an early start on my veggi garden with a few grow lamps in my basement each spring.
The energy is "free" because the furnace would run more without them anyway.
By the way, if you live in a cold environment and wondered why using LED bulbs didn't make any dent in your electricity bill, it's because your furnace runs more now without hot bulbs.
If your house is well-enough insulated for that to be true, you wouldn't need much heat in the winter anyway!
For most of us, forced air is far more efficient than lightbulbs spewing heat up into the attic.
Only if you rely on resistive heating, which most people don't, because it is ridiculously expensive.
Indeed; if you must use electricity to get heat you should use a heat pump which is 2-3x as efficient as resistive heating. A resistive load is 100% efficient. A heat pump moves heat from outside to inside the home so for the same electricity you can get 200-300% efficiency. (Make sure you get a scroll compressor though; the heat they put out is quite a bit warmer than the older piston-style compressors).
If you're thinking this sounds a lot like an air conditioner you are correct: Most residential heat pumps are just A/C units that can be run in reverse during the winter. They work well so long as the ambient temperature is high enough to allow the liquid refrigerant to evaporate (some low-temperature models can effectively heat a home when the outdoor air is around -10 deg C / 20 deg F).
Interesting idea. There's probably some clever software involved to schedule compute jobs onto the radiators; end users will want to regulate the heat output. Also, they must be using some rather beefy servers to generate an amount of heat comparable to a space heater (3kW).
In their FAQs they say they generate 1kW.
This is, to me, absolutely indistinguishable from an April fools' day joke.
You are not alone. How would they handle the dust and noise which comes so tightly with high performance servers?
Don't people pay good money for machines that generate white noise to help them sleep?
Not sure how I can spin the dust thing.
Servers don't generate dust, they collect it. Add a Hepa filter you have your spin.
We do not use a regular rack mounted blade, we have our own proprietary system that does not primarily use air to cool. It's completely quiet :)
network bandwidth is of course the biggest limiting factor. =/
Designing a data center with 1.5 mb/s upload between nodes would be.....not fun.
Excited to see if they can make a business out of this.
edit: also would suck that for half the year your data center gets turned off.
Not necessarily... it would be impractical to use these radiators as web servers, of course. But from their video, it seems their main application is number crunching. The nodes would download the input dataset, chew on it for a while, and upload the result some time later.
I think this is targeted to computation needs.
But on their registration page, they ask whereas you have fibre > 100Mbits/s.
Exactly we only use fiber, 10Gbps per house
More exciting than the free great is the idea of a reliable, high speed connection.
But I live in a South facing apartment in South Carolina, so I just bake something if my house is cold for a few hours.
What about wall-mounted out of the box bitcoin miners? Get around the physical access/security difficulties by making the homeowner own the machine and capture the benefit from its computation (rather than relying on network-based resource sharing). The economics might not really work out in the case of bitcoin but maybe there are other applications where the value of (computation + heat output) would make it worthwhile.
That doesn't really get around the problem; I could install a tap upstream of the miner, and interdict any blocks that get mined. It might even be worse than the current model, since the value of the work being done is completely exfiltratable (i.e. if I steal the mined block I get all the reward, whereas I might have to work to extract value from stolen computation results).
> if I steal the mined block I get all the reward
How would you steal a block? Changing the target of the payment requires recomputing the nonce.
The value is in the proof of work; just a very hard-to-compute number. All you need to do to steal the work is to prevent the server from sending the hash to the owner, and independently submit that proof of work elsewhere.
I've been heating my room by running stanford's folding at home. Silly but it actually works
home version of this http://www.geekwire.com/2014/amazon-gets-green-light-heat-ne...
this is such a good idea, it works well in the summer too
I suppose servers could be distributed worldwide, and run seasonally, except for the population imbalance between the northern and southern hemispheres.
We have a system to dump heat in the summer, and Dutch summers are sucky enough to not really make this a problem...