Microsoft to dump it's computers into the sea.

[bennyboy]

Working link here:

https://www.bbc.co.uk/news/technology-44368813

I must say I'm intrigued.

The point is that any savings you make scale up as you get much larger data farms. If you have to pump water in to cool your computers then it won't be as efficient because it takes energy to pump the water, which also produces heat. If you have it underwater then you don't need to pump any water or air and rely on tidal power instead.

The problem is though if that a single component fails then the entire server is unusable. They're saying that in an oxygen free environment this is less likely to happen but I'm sceptical because components don't just fail for environmental reasons but because they weren't built properly in the factory. So whether it actually saves money is unknown.

I don't think this is a big problem.  They're not running a single unit, like we do at home, and crossing their fingers that the thing doesn't go down while they're on holiday.

They can estimate the halflife, so to speak, of the units, and know what % of the racks will be functional after a given duration of time.

[johan]

The cost of space grade components to increase the reliability of their servers would be orders of magnitude more expensive than remaining cooled on land.

No need for space grade components. Weight won't be a factor like it is with objects bound for space so you can pack lots of extra backup units in there. It won't be Best Buy cheap but you should be able to do it for way less than the cost of space grade. My point was creating computer systems that work reliably hands off for extended periods of time is hardly uncharted territory.

[Anomalocaris]

Working link here:

https://www.bbc.co.uk/news/technology-44368813

I must say I'm intrigued.

The point is that any savings you make scale up as you get much larger data farms. If you have to pump water in to cool your computers then it won't be as efficient because it takes energy to pump the water, which also produces heat. If you have it underwater then you don't need to pump any water or air and rely on tidal power instead.

The problem is though if that a single component fails then the entire server is unusable. They're saying that in an oxygen free environment this is less likely to happen but I'm sceptical because components don't just fail for environmental reasons but because they weren't built properly in the factory. So whether it actually saves money is unknown.

What is the mean time between failure of the system inside each container? It is not that components can’t fail. It is that if meantime between failure is long, then the loss of a small percentage of these self-contained underwater computing capsules may not offset the cost of energy savings.

Also, why can’t the container not be pulled up for servicing?

[ignoramus]

They're fucking idiots.

Spend a quarter of the cost decking the roof with solar panels and chill those fuckers with free refrigeration!

[Fireball]

The cost of space grade components to increase the reliability of their servers would be orders of magnitude more expensive than remaining cooled on land.

No need for space grade components. Weight won't be a factor like it is with objects bound for space so you can pack lots of extra backup units in there. It won't be Best Buy cheap but you should be able to do it for way less than the cost of space grade. My point was creating computer systems that work reliably hands off for extended periods of time is hardly uncharted territory.

True, to the extent that they have the room for massive redundancy. I still think, like Anomolacaris, why can't they be retrieved? That seems seriously not thought out.

[Jackalope]

Working link here:

https://www.bbc.co.uk/news/technology-44368813

I must say I'm intrigued.

The point is that any savings you make scale up as you get much larger data farms. If you have to pump water in to cool your computers then it won't be as efficient because it takes energy to pump the water, which also produces heat. If you have it underwater then you don't need to pump any water or air and rely on tidal power instead.

The problem is though if that a single component fails then the entire server is unusable. They're saying that in an oxygen free environment this is less likely to happen but I'm sceptical because components don't just fail for environmental reasons but because they weren't built properly in the factory. So whether it actually saves money is unknown.

I don't think this is a big problem.  They're not running a single unit, like we do at home, and crossing their fingers that the thing doesn't go down while they're on holiday.

They can estimate the halflife, so to speak, of the units, and know what % of the racks will be functional after a given duration of time.

^^^

Anything like this is going to be both massively parallel and virtualized. It's a -lot- cheaper to scale horizontally than it is to harden individual systems. Servicability is not an in iinsurmountable problem if cost of deployment is low enough.

I'm a whole lot more concerned about environmental impact than feasibility.

[I_am_not_mafia]

I'm a whole lot more concerned about environmental impact than feasibility.

Yeah that was my worry as well.

Some of them are going to leak in big storms that are more likely given climate change. The sea is a corrosive and hostile environment at the best of times.

Saying that, the casing will need to be brought up again if only to replace all the computers once they become obsolete.

And saying that it will only increase the water temperature by a millionth of a degree is disingenuous because the whole point of the experiment is to see if it can scale up.

[Succubus]

So, obviously I have no idea..but I can offer two possibilities for their decision.  One, if the choice with water is between pumping and in-situ use, you always pick the latter option.  Second..if you do have to pick the former option..pumping saltwater is about the most costly thing imaginable..when it comes to pumping water.  Maintenance.  Better to pump freshwater in, mix salt in a tank, then discharge and do it again......but there are huge costs (and a bevy of laws) about that as well.

Pumping salt water is about the most costly thing imaginable?
Not really, we've been pumping sea water for years using stainless steel pumps and piping, oil tanker ballast and desalination plants are two that spring to mind.
Ok, for the scenario as described even if the pump/pipework needs to be replaced once a year it's still a lot cheaper than what MS propose.

[The Grand Nudger]

So, obviously I have no idea..but I can offer two possibilities for their decision.  One, if the choice with water is between pumping and in-situ use, you always pick the latter option.  Second..if you do have to pick the former option..pumping saltwater is about the most costly thing imaginable..when it comes to pumping water.  Maintenance.  Better to pump freshwater in, mix salt in a tank, then discharge and do it again......but there are huge costs (and a bevy of laws) about that as well.

Pumping salt water is about the most costly thing imaginable?

Yeah....when it comes to pumping water.

Not really, we've been pumping sea water for years using stainless steel pumps and piping, oil tanker ballast and desalination plants are two that spring to mind.
Ok, for the scenario as described even if the pump/pipework needs to be replaced once a year it's still a lot cheaper than what MS propose.

I didn;t say we couldn;t do it or it had never been done, only that its relatively expensive when compared to pumping freshwater and mixing a tank..or just using a body of salt water where it sits. The latter idea being what they;re looking into.

The bigger cost, though, bigger than pumping..especially with saltwater, is dumping. Discharge is heavily restricted (it;s 90% or more of the operational cost of most simple marine systems). So..you either hold the water indefinitely (and then you have the option of mixing a custom saline solution) or you pay to play. Off the cuf, another potential cost for a surface operation like what they;re imagining would be heat exchangers. They want to drop these things deep to make use of ambient temps..but water above ground is rarely as chilly as deep water...and the size of the containment means that it heats up a hell of alot faster.

From a technological standpoint...the idea has merit. Im not sure if it would be practical or environmentally sound, though, lol.

It sounds like something that a researcher dreamt up..waiting to be poo pood by an army of dead-eyed soulless engineers.

[rado84]

...........

They better strap a cannon to Bill Gates' feet and dump him into the sea cuz whatever his dirty fingers touch they ruin it.