What happens to the ecology if they move all the data-centers in the world underwater?
Works on paper, but in practice I think the kinds of designs currently being looked at need to be permanently sealed, which means every time something goes wrong or a server goes offline or shit glitches out, you've got to haul the whole goddamned thing back to the surface, cut it open, repair it, reseal it, and resubmerge it.Now... if you could figure out a way to do this and leave it accessible to a dive team for repairs, that might be an option, especially if they can be made modular for easy expansion. Stick it on the seabed in some chilly little lifeless stretch of ocean.
>>17017296The hardware cycle is too short for any of that to matter and most hardware failures in data centers are caused by air movement issues, which isn't an issue with submerged data centers as they're filled with an inert liquid. Most other issues are caused by having humans in the facility, which also isn't an issue with underwater data centers. Actual hardware component failure due to manufacturing defects are quite rare.If hardware in an underwater data center fails, those components are taken offline while everything else hums along. Components are not repaired. They just sit there, unused, until the whole thing is hauled up after two years when the hardware is obsolete. The plan then is for the contents to be gutted and the pressure vessel renewed for reuse with newer hardware. If implemented at scale, the cost of the pressure vessels might end up low enough that some companies simply abandon them when the hardware inside become obsolete. This will become a regulatory issue as some people will take issue with the ocean floor being littered with decaying tubes of obsolete computer hardware.>>17017291It shouldn't be a problem. The amount of heat produces is tiny compared to the thermal capacity of the ocean. Even for a local area, it's not going to have much impact. Maybe if you put a couple hundred thousand of them in a shallow bay, it could cause localized changes. The only real danger is if they're abandoned and over time start to leak. Even then, the damage will be localized. Given the history of human industrialization, someone, somewhere, will end up doing something stupid that causes some issues for that particular area but in the grand scheme of things, it will matter very little.
>>17017296microsoft did this and never had to pull them up for repairs because everything lasted 10x longer
>>17017367they're a great idea but the electricity to run the data center is still going to be huge. The cooling is only part of the problem, although having the ocean do that for free does reduce the energy required substantially.
>>17017367which part requires full data center submersion that isn't beaten in costs just by putting the heat exchanges into the sea? A nominal power draw for the pumps?Unless you are trying to preserve data through some kind of surface calamity, this really doesn't fit the bill.hmm land rights maybe.
>>17017405Heat exchange in a body of liquid is about the best case scenario for shedding heat. Water, heat, and electricity are the prime issues with data centers. Submersion addresses two of these. Electricity however is still a big issue and running power out to the sea floor, though not novel, isn't really done at this scale. I'm curious if the restarting of the Michigan Palisades nuclear plant will result with data containers in Lake Michigan. The contamination risk is quite low but the US and Canada tend to be conservative with what's allowed in the Great Lakes.
>>17017393If only there was some kind of reliable chain reaction that generated a lot of power.
