Microsoft sinks to new depths with underwater data centre experiment Microsoft has revealed that it's trialled an underwater data centre. Project Natick saw Redmond sink a capsule about a mile offshore from Seattle, an approach felt to have potential because about half of humanity lives on or near coastlines. Microsoft's thinking is that it's a good idea to put data centres near their users, …
If it can be scaled, the net affect should be a reduction in energy usage, and so 'should' help the environment.
Water (liquid) cooling is much more efficient than air cooling, as used in traditional data-centres etc. Make the cooling more efficient, and you reduce the energy wasted spinning fans, cooling air etc.
Bear in mind, if the servers were not in this tin box in the water, then they'd be sat in some other data-centre anyway.
"...
Re: Haven't they heard
It would also be ideal for growing an artificial reef on (cold water corals, like we get here in the UK), although I would stick it a bit deeper; 30m would be optimal...."
Warning...SCUBA joke (and not a great one, at that, but I couldn't resist)...
Make it 45m or deeper then you'd be able to justify getting technical divers in...
No one will dive to it. If they need to service it they'll raise it up. Probably isn't worth the cost of servicing, they'd just replace the whole thing as a unit.
It is the same theory as mega datacenters that are built as units that fit in shipping containers. They don't service those either, they just replace the container when enough stuff inside it has gone bad.
"...@TonyJ - divers
No one will dive to it. If they need to service it they'll raise it up. Probably isn't worth the cost of servicing, they'd just replace the whole thing as a unit.
It is the same theory as mega datacenters that are built as units that fit in shipping containers. They don't service those either, they just replace the container when enough stuff inside it has gone bad..."
You don't even get a joke when someone tells you it's a joke...? I even admitted it wasn't a very good one but hey don't let that stop you.
Seriously? It isn't as though the heat will just concentrate around it and boil the oceans when you approach it.
If every datacenter in the world was moved underwater it wouldn't cause ocean temperatures to rise far enough to be measurable. Do you have any concept of just how much water there is in the world's oceans?
Do you have any concept of just how much water there is in the world's oceans?
Exactly, and do you have any idea how much coal burning it would take to cause global warning?
I think the heat has value. We are just not very good at recovering energy, particularly if the output is variable.
Heat as a general concept has no value for energy recovery, it is only in the context of temperature gradient it has value. Unfortunately the temperature difference between inlet and outlet air in datacenters isn't very large so it is difficult for it to do work. If you could run datacenters hot enough to boil water then you could drive a turbine...
The heat has value if you want to heat buildings or make hot water, but they are already recovering that heat in many heating dominated climates. Google 'datacenter district heat'. It isn't of much use in say coastal California unless you could locate the world's largest laundromat next door to use up all that hot water.
Basically, seawater is an electrolyte, so any time you put any two even slightly different metals into it in reasonably close proximity you will get current flow. Which means a transmigration of metal ions from one pole to the other. A reason for large metal boats not lasting for ever.
Though MS didn't get where it is today by being completely stupid, although some stupidity has had a part to play. I am sure they have enough smarts to avail themselves of the latest materials technology to make a well sealed vessel. I think 5 years should be doable and if the things are designed to be recoverable, bringing them to the surface at server changing time for an overhaul would help then to last the twenty years.
Any research for ideas to improve things is better than none so kudos to them.
"I might be wrong, but if the servers are in the sea there is no need to pump any water anywhere so they save even more energy."
Question is how much would you save compared with the extra costs of engineering for such an environment, maintenance and upgrading.
I'd have thought installing in the base structure of a wind turbine would be a better bet.
"I might be wrong, but if the servers are in the sea there is no need to pump any water anywhere so they save even more energy."
I wonder how much water they need? There are many sites where it's suitable to simple divert water to where you need it and allow gravity to take it's course. Waterwheels have been around for quite a few years now. Two solutions arise out of that. Use the flow of water as is or put a waterwheel in to drive a pump if you really need to. I'm thinking overshot wheels here as the most efficient type but and undershot would drive a pump too and get a head of water to use as the coolant flow.
> exactly! Why not just pump seawater out the sea (or a river), pass it through some heat exchanges and then let it back out into the sea?
Because water is heavy and pumping it up from sea-level to a building above sea-level needs a lot of energy : far more than you save in reduced cooling costs
"
Because water is heavy and pumping it up from sea-level to a building above sea-level needs a lot of energy : far more than you save in reduced cooling costs
"
Not if you are circulating the water rather than releasing it at the surface. The weight of the water coming up is then exactly balanced by the weight of the water going down so the only energy you need is to replace that lost in friction of the moving water against the walls of the pipe.
If it did geothermal heat pumps would not work. The common variety pump what is essentially radiator fluid in a closed loop through the ground outside your house (or straight down under it if you have limited land) and are typically about several times more energy efficient than air to air heat pumps (which themselves are more efficient t than traditional heating and cooling) If you have a body of water nearby that is deep or fast moving enough it won't freeze all the down during the winter nor disappear during a drought, you run the pipes through it and it is even more efficient.
I was thinking the same thing as the OP, "why not put it on the coast" but quickly had the same thought as others did here - coastal property is very valuable. Once you move inland very far your savings would evaporate due the added distance making initial construction more expensive and reduction in efficiency.
Once you progress to the "sealed shipping container" building blocks for your data center which we have several years ago, dropping sealed containers on the ocean floor is at least worth exploring. I'd wonder how much a storm could disrupt things - you'd need it far enough away from shore that big waves won't affect it on the bottom. Things that can get churned up pretty good down there when seas get heavy so I wonder if 30 feet is deep enough.
"...Why not just pump seawater out the sea (or a river), pass it through some heat exchanges and then let it back out into the sea? Works pretty well for power stations and they need to get rid of quite a lot more heat than a data centre would ..."
Pumping seawater is a vicious thing, especially when heat is involved. The salt causes all sorts of problems when it solidifies. It can be done, of course, but it adds to the complexity and cost of any equipment you need to do it.
One of the reasons you see most power stations by rivers - fresh water.
Yes, typically you wouldn't pump seawater, but you pump water with some additives (to prevent corrosion) in a closed loop to dump their heat in the heat. The other problem with seawater is that it isn't just salt that's in it, there are a lot of living things. You suck in marine life you have all sorts of other problems, not to mention having environmentalists upset (they won't care about microscopic life, but suck in a few starfish and there's hell to pay!)
Getting rid of heat easily, maybe? Immersing everything below 10 meters/30 foot ( below the inversion line for a decent body of water. For the non-divers : nut-numbing BRRRRRR!!! 24/7/365 ) gives a nice and effective cooling medium. Good thermal capacity, very solid and stable temperature differential.
Of course the article also mentions the main problem with that in an oblique way: barnacles... Your radiators won't stay clean very long, so you'd have to come up with something clever to keep the goo and beasties off to keep things efficient. Just about everything that lives in the water loves a free lunch, and a nice warm surface will do just fine, thankyou.
One reason against "why not just put the data centre on the coast" is cost. The coastline is often premium residential property and not ideally suited to big ugly data centres ruining sea views.
It seems a bit of a wacky proposition, but these things are worth at least exploring.
If you build it on land or a pier then the DC is "dry". Storm surges would kill it. Putting them underwater means not having to deal with land weather. Also, no pumps, no cooling fans. It's not a bad idea but I imagine there's still a lot of engineering to be sorted out.
Call me a suspicious and cynical old sod if you like, but I wonder if someone in M$ is thinking "offshore" as in "outside of territorial limits" and, by implication, outside of NSA control? A marine based power source, and a satellite based coms system with a buoy on the surface would complete the picture. And should be perfectly feasible with today's technology.
Nah - can't see who'd gain from that.....
The NSA had the Navy tap Soviet submarine phone lines back in the 1960ies - AFAIK they still have subs and divers. A very interesting book on the subject is 'Blind Man's Bluff'
https://en.wikipedia.org/wiki/Blind_Man's_Bluff:_The_Untold_Story_of_American_Submarine_Espionage
(sorry about the source, but the book is good)
> Microsoft imagines it might be possible to power sunken data centres with wave energy
Easier would be to locate these pods within existing off-shore wind farms. Electrical power is local, cabling routes to shore are already established and shipping and fishing access is already restricted so no extra inconvenience on that front.
Put the 20kw server in my basement. In Canada.
You pay the server power bill. I get free heat.
I get to sleep with the window open, even in February.
In summer, we might need to run a pipe from the lake.
Summer is often one week in late-July. Rest of year, air cooling.
I've offered to store old fuel-rods from nuclear power stations.
Tens of kilowatts of free heat in winter, worth the neutron flux.
Data centres could be submerged in the middle of large rivers, such as the Thames. Natural flow of fresh cold water, close to large conurbation and business district, and good power availability. I am sure there are other large cities with commercial centres straddling rivers, such as Koln, Budapest, etc. etc.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
