"We are in talks with two of the well-known hyperscalers, though, so it's still a little up in the air," Reynolds said.
Surely it should be "down in the water"?
A company called Subsea Cloud is planning to have a commercially available undersea datacenter operating off the coast of the US before the end of 2022, with other deployments planned for the Gulf of Mexico and the North Sea. Subsea, which says it has already deployed its technology with "a friendly government faction," plans …
so do the techs that service the unit have to have a current diving qualification card?
/me hears Jethro Tull doing "Aqualung" in my head
also was thinking it might be easier to put it in Lake Michigan (near Chicago) or in the lake behind Hoover Dam (somewhat near Las Vegas). The whole idea of cooling by immersing in water sounds like high potential for efficiency in ANY large lake, natural or man-made.
In my experience the number one cause of outages in data centres is well-intentioned but fat-fingered IT guys like me, tripping over cables, bumping buttons, etc, "Who, Me?" style.
I'm not even slightly surprised that a datacentre running with minimal human interference would typically tick along just fine.
In my semi-recent 9 year tenure at a corporate data center with a few thousand servers, whenever users called about what seemed like a server just dropping off line: "Was anyone in Pod X row Y recently?"
The whole data center staff was pinged immediately. We had to first prove none of us were guilty of a cable(s) faux pas THEN look to triage the issue.
...And maybe 10% of the time, we were the problem.
Yes, I did wonder about the article stating "less complexities in terms of deployment and maintenance".
Deployment, maybe, but maintenance? I'd imagine the only option is that customers can only rent CPU and storage capacity in a pre-filled container, which will have spare capacity to allow for failed units. I can't imagine putting your own kit in one of these simply because you won't be sending anyone down to replace a card, a drive, a PSU, a system board or backplane etc. You either pre-equip the racks with hot spares or when it fails it gets expensive to repair. "Cloud" rental, not Co-Lo.
"According to Subsea, customers can schedule periodic maintenance, including server replacement, and the company says that would take 4-16 hours for a team to get to the site, bring up the required pod(s), and replace any equipment."
I wonder if "schedule periodic maintenance" is on-demand or on a set schedule? I mean, 800 servers could potentially be 800 clients, which, if on-demand, could mean the pod spends as much time on the surface in dry-dock getting hardware swapped out as it does submerged. Or is each pod dedicated to a single customer? That might make more sense, but then it's potentially a bunch of less-than-optimal pods hanging around down there.
Other than exploiting the ambiguity of offshore real estate regulations in many countries.
Shallow fresh water, on land, where you won't be scraping crusty barnacles and jellyfish out of your heat exchangers, and technicians can get in the module from a hatch instead of having to winch something 160 fathoms off the seabed. Or deal with someone dragging a naval anchor over it. In fact why even have it in the water when having a shallow cooling pool under it would suffice?
Is this another one of those dingy Crypto-Libertarian free state ideas like the Trojan iceberg? At least our Navy SEALs and the CIA can log some dive time installing taps on them on the sea floor when no one is watching...
existing dams and the U.S. Great Lakes are good categories, so long as unfriendly gummints and wacky enviro-activists don't drag the process out to infinity with continuous lawsuits, protests, cancel culture, etc..
If things get to THAT point, sea water (vs fresh) is less of a problem. It would be a matter of doing anything at ALL. And that goes DOUBLE for anything attempted in Cali-F-U . (these are very realistic concerns; years of permitting, protests, legal challenges, etc. can easily kill ANY new construction -project by draining its funds before a single concrete slab gets poured or brick gets laid)
Worth pointing out, barnacles are found close to the surface where the plankton is. So down where it is dark (below about 75 meters), it would be less of a problem. In fact, MOST of the sea growth that can clog up cooling systems is generally found close to the surface.
Also worth pointing out, water pressure increases about 1psi per 2 feet. I do not know what that is in metric, as I observed it while I was on a sub in the U.S. Navy where everything was still in feet and psi (etc.). But the deeper you go the stronger the vessel needs to be. You COULD pressurize it with air, but you do not want the equivalent of a diesel engine at depth going off when pressure meets anything with hydrocarbons and electricity, so it would have to be something inert (like Nitrogen gas). That adds more expense. It makes shallower water look a lot better.
> Also worth pointing out, water pressure increases about 1psi per 2 feet...
From the article: Inside, the servers are also immersed in a dielectric coolant, which conducts heat but not electricity.
I.e. the pods are filled with liquid which is as incompressible as the water outside, so no need for any diesel pumps to pressurise them, just a simple equalisation mechanism (squeezy bag).
Though you do want to be careful of any gas pockets within the kit itself, so SSDs instead of helium-filled HDDs perhaps.
Probably because sea-levels are rising, and are not subject to the vagaries of drought/climate change. Go see Lake Mead and the mafia dumped bodies - summer 2022.
“The savings are the result of a smaller bill of materials, and less complexities in terms of deployment and maintenance," Reynolds told us. "It's complex and costly to put in the infrastructure in metropolitan areas, and in rural areas too: there are land rights and permits to consider and labor is slower and can be more expensive."
… surely it’s costly - massively expensive downside - to put it in a permanently leak-proof ‘server submarine’ too ?? With occasional need to disrupt/bring up for servicing. Apple warranty weasel on IP68 water resistance on an iPhone 11 so this needs to have great seals !!!
Aside from the issue noted above with lakes having a habit of drying up, dilution. Things like power stations use rivers and the sea, rather than lakes, for cooling because you need flowing water or a large enough heat sink, otherwise said heat sink heats up too much and stops being a sink. And even then they need to be careful not to heat the local environment up too fast for the flow to disperse so they don't kill everything trying to live in the local water. A shallow cooling pool is not going to cut it for cooling a 1 MW heat source. Put a few of them together, and even a decent size lake is going to end up being completely sterilised of all life more complex than algae.
On the other hand, lifting things out of the sea isn't all that complicated. It mostly causes issues when the thing being lifted wasn't supposed to sink in the first place. A shipping container deliberately put in place with lifting hoops just needs a fairly small boat with a couple of chains with hooks on the end. Slightly more complicated than having it on land, but as long as the savings in cooling are more than the extra cost, what's the problem? More difficult to get to on short notice as well I guess, but that's why it's being aimed at big cloudy types and not people looking to site a single server.
If the only reason for dunking it in the sea is to keep it cool, why not just winch it off the end of a pier? Then you could get to it by car much more quickly.
Or what about a lake? Plenty of deep, cold lakes about in mountainous regions. Have they thought this through properly? Way out at sea, there's also a risk that your datacentre might be hoisted and removed by some other counry's submarine as well.
How many deep cold mountain lakes have the kind of fiber and power connectivity that would be required? Both tend to be present on the shores of major cities.
I assume they are kept close enough to shore that they don't have to worry about submarines, or fishing boats. In places where undersea seawalls are used to help reduce beach erosion this would work as well as sinking old containers full of rocks which some places already do...
>Have they thought this through properly?
If they haven't got plans for cloud data centre sized deployments, I suggest not.
As has been noted, Microsoft and others have demonstrated the viability of submerging a single container of servers, but stacked containers or scaled out to cloud data centre size - circa 100 containers, I suggest not.
I can see that the Coast Guard or Navy might keep foreign marauders away from phishing data containers where they lie within coastal zones. And I guess there are legal precedents for keeping thieves from open water salmon farms, based on historical fishery law. But anything unmanned left lying on the seabed looks awfully much like flotsam to me, free for anyone to salvage and take posession of. I'm thinking that harvesting open water wind turbines isn't more popular mostly because they seem to be rather firmly anchored into the ground.
And then, once this becomes mainstream, how do you govern and manage the space? Evidently coastal property doesn't automatically extend into the water, in many more civilized places the coast itself is public land and on top the sea is both shifty and three-dimensional: top positions could wind up getting slow cooked from ground feeders in a race to the [ocean] bottom.
I've always been wondering about the legalities around marine cables, especially since there seem to be specialists who dig them up and ...steal[??] them: how can they be stolen, when they have been abandoned in international waters? How does the repair ship even stop a theft in progress, other than by superior force?
And should there be cabling conflicts due to cross-overs and repairs required: how is that being managed?
Once you add server farms to sea cables, complexity is sure to follow them closely.
Concentrated mass deployments could imply heat pollution, which is already a big issue at many of the power plants that use lakes and rivers for cooling. That's why a combination of tidal power generators with a DC (data centre) add-on seem much ...cooler, especially with DC (direct circuit) power.
My biggest worry is that those open sea data center containers will "unfortunately" get lost at sea, once they've outlived their usefulness (or the owning shell company has sunk).
heat is heat , the planet's a closed system. heating the oceans is no better and perhaps nastier than heating the air. remember that the sea harbors all kinds of life , mammals , fish etc and they're relying on normal water temperatures ( regional ) to survive. Heating it is not a good idea.
But it's got to be greener to contribute to ocean warming directly than via atmospheric CO2 emissions... :)
The trouble is, a single container will probably have little effect other than on what lives and grows directly on it. But scale out to a hundred or so containers together (100 x 800 = 80,000 servers) and things will be very different.
I am no marine scientist, but i suspect many ocean dwellers (fishes, corals etc.) are sensitive to the temperature of the sea water (maybe big change will happen even if it is +/- 1 degC, like massive death), and if this below-sea data centers proves to work then there will be a rush to put these heaters under the sea(much lower cooling cost! what's not to love!) and a catastrophe will happen to those ocean dwellers and the surrounding ecosystem.
Hope I am wrong.
So this bloody Bstd will be heating the oceans at 700 ft. Doubt me and you doubt the second law of Thermodynamics. I don't mind. The ocean is a big place. However when you advertise the greenhouse benefits to secure "environmentally conscious" investors' money, and idiot money from the Government, perhaps some truth in advertising is warranted. Then again, why bother? The only downsides I see at this point are pressures and repairs. I assume they are going to put some kind of GAS or air; pray to tell not Carbon Dioxide in this thing. A few bubbles of that toxic stuff and you will be decried as Satan of the Seas or worse the thief that stole Greta's adolescence.
So let's talk repairs. Will a tech be dropping down in a submarine to fix this thing, excuse me, the data centre? Will they recover it, or simply attach a new one and release the old one into the shipping lanes? It seems a bit elaborate to me just to avoid paying for some air conditioners.
I sort of think they fill the entire container with mineral oil (or whatever their immersion fluid is). I couldn't find specific evidence one way or another, but reading through their site, the clues seemed to point to the entire thing being full of fluid. I could well be wrong, though, and may have missed a page where they state it clearly.
Yeah, some sort of fluid.
Scuba-diving maintenance engineer just swims down, opens the fluid-lock door, goes in and does the fix.
For a large installation might even be worth engineers doing shifts in an attached pressure chamber to save time.
Literally from this article (admittedly near the bottom, so perhaps your attention span was too short to catch it):
"The viability of underwater datacenters has already been demonstrated by Microsoft, which has deployed several over the past decade as part of its Project Natick experiment."
Instead of "Cloud Computing" we can have "Protean Processing" (or "Pelagic Processing", but the shape-shifting aspect seems appropriate).
For "Cloud Storage", maybe "Salacial Storage" (as in 'of Salacia', careful not to confuse that with "Salacious Storage", as if) or maybe simply "Doris Data" (not to be confused with "Doric Data", which would be a spreadsheet).
Full fathom five thy father lies;
Of his bones are coral made;
Those are pearls that were his eyes:
Nothing of him that doth fade,
But doth suffer a sea-change
Into something rich and strange.
Sea-nymphs hourly ring his knell:
Hark! now I hear them,—ding-dong, bell.