Follow Slashdot blog updates by subscribing to our blog RSS feed
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
The datacenter modules are more efficient and less error prone because no one installed windows on them.
It seems like they would get about 90+% of the savings by just leaving the pod on land and using an underwater heat exchanger instead. The same amount of heat will be produced by the servers, and a small motor to pump the radiator fluid would remain above water and air cooled. I don't see any other CO2 savings by immersing the whole pod at a depth where it isn't a hazard to shipping lanes.
Unlike the fully submerged model, if someone needs to work on their servers, you don't have to take the entire server pod offline. If cooling for each rack or 4U block can be isolated for maintenance, the many of the other 799 servers can continue to operate.
I know many large data centers are currently being cooled by either using treated drinking water, or using nearby rivers as huge heat sinks. Neither option is garnering them many supporters. https://time.com/5814276/googl... [time.com]
The standard /. armchair instant engineer analysis -- I just heard of this but I know its dumb because...
There are several advantages mentioned, among them not needing expensive coastal land to get the data center near the ocean for cooling, or near the city for low latency. And you can make it not a hazard for shipping lanes by not putting it in shipping lanes. They are called "lanes" because the restrict ships to specific channels. Then too - the maximum draft of a container ship (a very large one) is 16
It seems like they would get about 90+% of the savings by just leaving the pod on land and using an underwater heat exchanger instead.
It seems like they would get about 90+% of the savings by just leaving the pod on land and using an underwater heat exchanger instead.
Waterfront property is not that easy / cheap to get as just dumping something underwater. Also then you need infrastructure to move water which costs money. There's some real benefits to simple submersion.
Plenty of downsides too, but I'd like to think a company investing many millions of dollars actually did a Cost-Benefit analysis, and while we have nothing good to say about "UX Experience Managers" at Microsoft, the entire company isn't staffed by idiots.
As you point out, the green lobby is very worried about your exit water being a quarter of a degree hotter than its intake and thus on-shore, cost and energy effective methods are being opposed over these which because they donâ(TM)t use conventional ac will now be designated green.
This can be resolved with shell-scripting!
Homer, that's your solution to everything: to move under the sea! It's not going to happen!
Homer, that's your solution to everything: to move under the sea! It's not going to happen!
You now have the Little Mermaid song running through your head. Happy Friday.
Surely it would be swimming through my head.
We see no commercial offering from them yet, so perhaps it wasn't quite as great of a success as they had hoped.
We see no commercial offering from them yet, so perhaps it wasn't quite as great of a success as they had hoped.
You typically don't go from such an outlandish concept pilot to full commercial in a short period of time, especially when you promise your customers certain reliability. I suspect it may have been a roaring success. Mind you COVID was a bit of a distraction the past few years too.
1 year? What sort of shoddy outfit are you running?
Within 6 months expect to a hard requirement of,
"10+ years undersea pod management experience in 10 or more oceans".
Knowledge in Shell scripting and C Shark required!
> I think we can write "Phishing attack" with an F at some point.
Oh it'll get F'd alright.
Actually, if they use lakes this could be a problem. Not sure. It's been a problem with nuclear reactors. Of course, it would take a really LARGE data "pod" to be the equivalent of a nuclear reactor.
Yes, and lakes vary a lot in size. Also sensitivity. E.g. in some lakes disrupting the thermocline from forming in winter could kill various kinds of animals, especially the extremely small ones near the base of the food chain.
I can't see any of the great lakes having a problem with this, or any of the larger rivers. (And only the larger rivers are likely to be deep enough.) But with smaller lakes that don't have a lot of flow through ... yeah, it could be a problem even for a bunch of data pods.
So maybe it would be simpler to leave the servers on land or at least house them in a river, deep lake or quarry where some of the issues could be mitigated.
I was thinking along the same lines. Thinking HDD replacements, you would have to pull the whole thing up.
They must deploy these things with a ton of extra redundancy and only haul them up on pre-arranged maintenance schedules.
No you scale horizontally and don't care about any particular server. When, say, 30% of them have failed in a container you move services to a different container and do maintenance on the idle one.
Plus it sounds as if you're tied to whatever technology was state of the art at the time you sank the pod. While the pace of CPU improvements has slowed somewhat, most other technologies are still rapidly improving, especially storage and network bandwidth.
It's an interesting idea, and I'd be curious to know how they've addressed these issues. Presumably marking a pod as obsolete after five years doesn't really fit the green ambitions of the project.
If a server goes down or requires on-site maintenance you're basically fucked. And is the whole pod suppose to be risen to the surface to just to take out one bad card or whatever? Seems inconvenient.
If a server goes down or requires on-site maintenance you're basically fucked. And is the whole pod suppose to be risen to the surface to just to take out one bad card or whatever? Seems inconvenient.
You are inadvertently making a point for putting the pod underwater (though simply locking the door can accomplish the same thing).
In a modern high up-time data center you do not want techs stumbling and fumbling through the racks to swap out a single card, because having people moving around in the data center, is a key problem in getting those high nines. Modern centers are "lights-out" no access places except for rare occasions of major system upgrades. Workloads are distributed across the entire server farm and are moved from box to box automatically as needed. The idea of a single box committed to one application that it tied to it is obsolete and has been for nigh upon twenty years now when server virtualization became common.
You'd be amazed at the number of people who when you ask them "What's heavier, a pound of feathers, or a pound of lead" actually stop and think about it.
...Underwater Basket Weaving will finally pay off. Toldja, Mom!
Wouldn't something like Lake Michigan be colder and less corrosive?
But then it's in Michigan...
Corrosion isn't some unknown or unsolvable problem. It's just a simple engineering input which needs to be managed. Hell I just got off a floating vessel which has been stationary and ankered in the ocean, never removed for over 25 years.
Our sprinkler system recently turned our datacenter into an underwater datacenter, but that didn't really make the news. Well, beyond our customers, that is...
This screams "investor trap" to me.
Under the sea no one can hear you scream. Except maybe the whales, but they are becoming extinct.
Why? MS has been investigating this for years, and even done actual trials to prove the technology.
Yes, you've got it. They say that, core for core, it will cost 1/10th the price of a traditional data center since they don't have e.g. the costs of HVAC, human factors, and greatly reduced real estate costs.
I'm skeptical of this statement because of the lack of COTS hardware rated for dielectric immersion. They might hit it eventually, but right now they're going to be hand building this kit
...this business will go under.
.. it will go sink like a rock....
Seriously, did anyone else read this as "Underwear data-center" first time you saw it?
I really hope it's not IT administrators underwear you are talking about.
I can see the ads now... Must have prior data center experience and current Scuba certification.
they are trying to solve. I did some analysis and data modeling for a geothermal power plant, client work, close to 15 years ago. I kept an eye on the prototypes and how it progressed. They had to change their business model because of the cooling problem. The issue was that, like in a previous comment, they pumped the excess heat out into a local river (it's really big isn't it?) and unsurprisingly fish started dying and (poisonous) algae grew at the site. ... Is the Pacific Ocean big enough to absorb ANY amount of heat we need to get rid of? It seems doubtful.
they are trying to solve. I did some analysis and data modeling for a geothermal power plant, client work, close to 15 years ago. I kept an eye on the prototypes and how it progressed. They had to change their business model because of the cooling problem. The issue was that, like in a previous comment, they pumped the excess heat out into a local river (it's really big isn't it?) and unsurprisingly fish started dying and (poisonous) algae grew at the site. ... Is the Pacific Ocean big enough to absorb ANY amount of heat we need to get rid of? It seems doubtful.
Simply put, yes. The amount of heat we humans can directly generate could easily be dissipated by the Pacific Ocean. The problem behind global warming is that we have accidentally leveraged a MUCH larger source of heat--the sun.
This particular test is in the Strait of Juan de Fuca, which supplies far more water flow (est 20-30 times more) than all the earth's rivers combined (https://en.wikipedia.org/wiki/Juan_de_Fuca_Channel#The_amount_of_water_in_the_channel). So as long as they place the containers in a
1 MW of heat dispersed over an ocean is insignificant. However, there is no heat dispersal mechanism, so the temperature rise near the container should be quite high, with some gradient with increasing distance from the container. 1 MW in a local area is likely to affect/kill most animal and plant life near the container, with additional consequences for the down-current environment.
Subsea Cloud: Hi I'd like a business loan. Loan Officer: Tell me about your company. Subsea Cloud: Well... It's going to be underwater pretty soon... Loan Officer: Sorry, we can't help unless you have positive assets. Subsea Cloud: No. I mean the property will be underwater. Loan Officer: Sorry, we can't help if you're property is flooded. Do you have insurance? ...
Subsea Cloud: Hi I'd like a business loan. Loan Officer: Tell me about your company. Subsea Cloud: Well... It's going to be underwater pretty soon... Loan Officer: Sorry, we can't help unless you have positive assets. Subsea Cloud: No. I mean the property will be underwater. Loan Officer: Sorry, we can't help if you're property is flooded. Do you have insurance? ...
lowering latency by allowing the datacenter to be located closer to metropolitan areas,
How long before we have undersea data centers actually build over undersea fiber optic cables and tapping directly in?
I like this thing from an environmental perspective, and it really does seem like it will have some energy advantages. However, what about physical security?
In the developed world, there's no real concern that someone will sabotage- either for political reasons, religious reasons, or even just for the lulz- a datacenter (or any other building with a lot of valuable assets in it). They are reasonably large and secure, and like everything else, reasonably surveilled. Someone trying to damage them would hav
There may be more comments in this discussion. Without JavaScript enabled, you might want to turn on Classic Discussion System in your preferences instead.
The T-Mobile/Sprint Merger Hasn't Created Jobs -- It's Cut Thousands
Royal Caribbean Will Equip All Its Cruise Ships With Starlink Internet
Those who do not understand Unix are condemned to reinvent it, poorly. - Henry Spencer, University of Toronto Unix hack