And this is where the idea originated...
Your next data centre could be an aluminium box filled with a handful of servers swimming in oil. That's the vision El Reg today beheld at the OpenStack Summit in Sydney, as advanced by a pair of European companies. One of the companies is called Horizon Computing, which has created the “RuggedPOD”, a 70cm x 70cm x 70cm box …
Come on who is not thinking this.
Joking aside the question is how the costs stack up.
Dense, but highly cooled --> fast
Not so dense, passive cooled --> cheap
Given that Google have talked about (have?) locating server bunkers in Iceland to side step the cooling bill this might not be so stupid.
However de-greasing the boards for trouble shooting purposes without CFC solvents could get messy.
A bit of vinegar with the oil could make a nice simple dressing for a salad.
Not so dense, passive cooled --> cheap
What I'm not seeing is what this "saves" besides the running cost of the server fans. It doesn't automagically make the heat vanish, it just ports it from the components to the room passively, so with anything resembling a conventional DC you still need the same CRAC capacity as before.
Presumably that's where the outdoor bit comes in. They envisage server farms as literal "farms" - open fields of 21st century menhirs exposed to the elements. I suspect that's going to be a tough sell, even if they have already accounted for heat sink fins clogged with seagull crap and all the other wildcard issues.
Yep, nothing new under the sun ...
The downside is fixing the things as the boards are slimy to work with after soaking ... Modern SSD makes the storage viable for submersion too (which spinning rust didn't like).
The two approaches were simple submersion relying on natural convection for lower power silent systems, and pumped coolant for high power boxes which was better but in itself obviously requires more power and loses efficiency. I believe some of Chinese bitcoin mining racks are pumped coolant systems.
Didn't Thinking Machines have a liquid cooled supercomputer 30+ years ago?
ISTM the idea is nothing new and doesn't seem to have changed the world.
As for placing this outside. That's fine until the winter arrives and every bug, insect, and crawly thing is attracted to the nice warm environment and starts to make a nest / cocoon / web in those nice, accessible cooling fins. Followed quickly by all the birds and animals that eat those things.
It would probably be better to drop these boxes in a lake somewhere and use the warmed water for fish farming or summat. That would make a service tech's job even more interesting. "Yes, I'm a qualified Microsoft engineer and SCUBA diver!"
"It would probably be better to drop these boxes in a lake somewhere and use the warmed water for fish farming or summat."
All that happens then is that you get the aquatic version of "every bug/crawly thing - and the things that eat them" taking up residence and clogging the cooling.
Transformer oil will quite happily dissolve or swell some types of plastic and rubber. This could be a problem if nobody told the ITX board builders that their boards will be dunked in oil.
The dielectric constant would be higher too, meaning the impedance of tracks in contact with the oil will change - is this significant enough to matter? Who knows?
They need to have expansion space for the oil and some removal of water, silica gel usually, like one does with a normal power transformer.
I kinda suspect that the "certified for oil" ITX-boards will be more than expensive enough to eat any savings on "the environment" - or - one does the usual IT QA process: Slap it together and fire that sucker up, if it runs long enough to bore the tester, it's Golden.
Standard transformer oil will dissolve many of the plastics and elastomers found in electronic circuitry. My company makes specialty synthetic dielectric coolants for immersion cooling of servers, batteries and motors. Electricity savings are >70% over standard air cooling. www.engineeredfluids.com
30+ years ago I was working on radar systems that used immersed circuitry in silicone oil and it was old technology in those days (just wish they hadn't made the crates so deep - working on them in winter having to strip to the waist to undo screws etc at the bottom of it - flipping freezing in an unheated bay of an aircraft).
Heat exchangers could put the radiated heat to some good use, there's a server farm in Sweden where the exchanged heat is sold off cheap to heat a local housing estate of a Cpl of dozen houses.
If you wait logon enough things come back round (sometimes to haunt you)
<quote>"installs servers that look like radiators into domiciles. The residents get free heat"</quote>
Nothing new here! Many a radio or TV station located in the colder climates often used 'heat exchange' to dump the waste heat from their tube ('valve' for those of you on that side of the pond) transmitters into their studio facilities during the winter. Hey it is free heat after all!!!!
erm I think that's actually Bowmore distillery that does that. The distillery is right next to the only high school on the island and is in the biggest village. I'd be very surprised if it was Lagavulin as you would be lucky if there are more than a dozen or so locals in the vicinity...
Who the hell would trust these things to stay put if left outside somewhere? If you left them in a literal field any cost savings are going to be eaten by trying to stop any and all tea leaves from nicking the kit.
That said you could always stick it on the roof of a building, that would at least make it more of a challenge but then the afore mentioned bird crap/insect infestation/rodents looking for somewhere warm might not go away.
That all said - I can think of uses for it, ships would be able to make use of such boxes since it would allow more computing power for less cooling, the army too might be quite happy with boxes like this (just make sure they're up to being dropped... alot, no offence to army boys but if i'm carrying a heavy server and get shelled... I'm dropping it and running) there'd also be a use for it in large industrial sites since it's less likely to pull whatevers floating in the atmosphere inside itself and die from some random particles in the air (metal filing, pollution, air bourne acids, etc etc).
I just don't think this is going to be a mainstream product anytime soon.
I just hope they don't use PCB's to cool their PCB's
"Some PCBs share a structural similarity and toxic mode of action with dioxin. Other toxic effects such as endocrine disruption (notably blocking of thyroid system functioning) and neurotoxicity are known"
Leaving them outside is impractical for a variety of reasons, as stated above, and what about leakage? Who's responsible for groundwater contamination? I know, I know in the USA and Canada it'll be the taxpayer, but in many more responsible countries it could be the company that deployed it to its location. I'd think that the insurance to coverage that eventuality would raise the cost.
You mean the company who deployed it to the location and which conventiently goes bust shortly before the clean up is due to start?
If we're using technology like this companies should pay a levy for clean up etc which is then refunded with interest IF the money isn't used to clean up their mess once the site closes.
As someone with a bit of a sideline in rural IT, I have encountered "open-air" problems you just wouldn't believe. Rodents, insects, birds, curious big mammals (including thieves) --- and that's before the weather. I mean you can rainproof all you like but a sudden drop in the external temp often causes internal condensation. And dirt. Where does all that dirt come from? And how did it get in here?
Hydrocarbons eat plastics. Silicone oils still damage plastics, just much more slowly. Flurocarbons... are actually very safe for electronics, but are also crazy-expensive. Regardless, the only way to be sure electronics are going to endure for years of service is to run them through years of service.
I can see possible niches for applications where space is tight and air quality is appalling, like industrial or military use. No need to worry about dust build-up.
I wonder how these would go on the surface of Mars. Go plonking them down outside, connect a cable that supplies data and power, instant processing power without wasted power draw from cooling systems, low/no maintenance, and you could even reprocess them once they get to a certain failure rate eg: You have two units with 50% failure, take them inside, drain the oil, swap the good bits from one into the other, then you have one unit back at 100%, and an empty chassis with oil ready for a shipment of upgraded computers to arrive from Earth
Immersion cooling in a dielectric liquid is the next step in data center cooling, as the electricity required for the aircon comprises most of the data center's power use. Immersion cooling reduces this by up to 70%. It can take several form factors; this is one, but also as seen at liquidcoolsolutions.com and midasgt.com. While ordinary mineral oil will work short term as a dielectric oil, it has problems with material compatibility, oxidation resistance and zinc whisker growth. As noted above, fluorinated fluids work well, but they're hugely expensive and are usually used in two-phase systems where the fluid actually boils and is condensed and pumped back to the components. An engineered dielectric coolant that's specially made for this application is the way to go.
Unfortunately the solution as-is doesn't help with that, because it still depends on air to move the heat from the liquid cooling enclosure and out of the room. Perhaps if it included a water-cooled heat exchanger. Easy enough to imagine a situation where, rather than huge air systems, each enclosure has two hoses connected to a ceiling manifold.
Biting the hand that feeds IT © 1998–2019