Slashdot Mirror
Microsoft Serves Cloud From the Sea Bed (datacenterdynamics.com)
judgecorp writes: A Microsoft Research project to run a data center underwater was so successful the team actually delivered commercial Azure cloud services from the module, which was 1km off the US Pacific coast for three months. The vessel, dubbed Leona Philpot after a Halo character, is a proof of concept for Project Natick, which proposes small data centers that could be submerged for five years or more, serving coastal communities.
Sorry for hijacking this article, but I would like to say that since the latest takeover, we have seen much higher quality articles than we saw pre-takeover. The articles all appear to follow the "News for Nerds. News that matters." tagline that Slashdot used to follow. It is early, but I am cautiously optimistic that things are getting better.
Keep up the good work.
That's your answer to everything.
Those who do not learn from commit history are doomed to regress it.
...but I don't see any Windows on that capsule.
-- Thou hast strayed far from the path of the Avatar.
Maintenance must be a killer... Having to dive to fix a problem. I am not even making fun of Microsoft track record of less than stellar reliability to make 5 years of uptime seem possible.
But connections to the systems, Cable get corroded or broken.
Pirates you have millions of dollars of equipment under the sea mostly unguarded. If they may want to bring it up to steal and sell the hardware... Or they could hack into it the hard way (To get information from it)
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
in this sort of virtualizarion hosting environment, you don't fix it. If it can't be resolved remotely you simply take it offline and permanently shut it down, much like bad sectors on a hard disk or SSD get remapped and excluded.
So their portable data center is about the size of a container. Why not put it on dry land? Certainly renting ground the size of a container from someone has to be cheaper than running undersea cables. This seems like a stunt, not a business plan.
But this is a sunk cost.
When the servers crash and display a blue screen it's not so evident anymore.
Cold ocean water transfers heat away from the container much more efficiently than warm air would. A cable might be more expensive than rent, but is it more expensive than rent + air conditioning?
"[Regarding the 'cloud,'] ownership was what made America different than Russia." -- Woz
Microsoft has put a video on Youtube, and a new blog [post about Natick today. They are both linked from my article. http://www.datacenterdynamics.... The Youtube video is here https://www.youtube.com/watch?... Peter Judge
I wonder if Microsoft is doing this for legal reasons. If the go far enough from the shore, they're in international waters, and are not subject to national laws anymore.
This is really old news. Using deep-ocean installations to nominally negate the costs of cooling in data centers had been around forever.
And energy-harvesting by use of undersea currents, tidal motions, or hydrothermal vents has been around forever, too. (Geothermal energy, anyone?)
This article has nothing new, but its author's suggestion that co-locating the 'pod'-type data centers near undersea thermal-emission sites is flat-out stupid. An umbilicus to land, eventually to an internet trunk-line is required. We can pipe around photons and electrons with ease. So why, oh why, was the writer forced to fill column-space with this nit-witted statement?
There are plenty of reasons to emplace various things at-depth in our oceans, simply for the heat-removal aspect alone. Below 400 m it's all pretty much below -3C. Using service-life maintenance-free modules is a great idea —It is not new.
Soo.....what's the benefit of having it below water, offshore?
I'm assuming, heat dissipation. I have a thought: could the column of steam rising out of the sea above a really big server farm be usable, given an onshore prevailing wind, to increase rainfall in dry places like California?
Stupid sexy Flanders.
It is measured in nautical miles = 1.15 mile = 1.852 km
By the way one knot = one nautical mile per hour.
Everything I write is lies, read between the lines.
Prefab means they can be deployed relatively quickly. You can prefab on land, but you still need to lay foundations... and these wouldn't be permanent; when you wanted them gone, they'd pack it up and remove it, leaving no trace.
Maybe not permanent, but designed for 20-year residence, with 5-year hardware rotations. It's enormously more expensive to build a watertight submarine than a waterproof house, so you're going to want substantial lifetime to recover that extra margin. Water-tight seals, corrosion resistance, pressure proof to say 3-4 atmospheres. The actual chamber isn't flooded, so there's still a big air space and a thick steel plate between your hot processor and the cool water.
I can see where the real estate to plant these things might be cheaper than downtown San Francisco. They're still going to need some kind of shore facility to house routers and power.. Nor do I think there are that many real-world installations that regularly go 5 years without any hardware maintenance.
Seems like a neat toy. Or a publicity stunt. I'm glad someone has so much cash lying around that they can investigate wholly impractical applications, but no one who knows anything about water is likely to go anywhere near this. If you need fast and impermanent, put a shipping container in a parking lot. If you want passive cooling, find a parking lot in Anchorage or Edmonton.
But this is a sunk cost.
Even so, on this whole investment, they must be... underwater.
The project is probably run by someone young, who is still... wet behind the ears.
Sorry. I'm just... fishing for karma. I'll say goodbye now... *waves*
You shall see a cow on the roof of a cotton house.
We can expect more 'accidental' NSA anchor drops on cables if this ramps up into production.