15
votes
Space data centres: ‘A figment of the imagination’ but one that could make Europe a space leader
Link information
This data is scraped automatically and may be incorrect.
- Title
- Why Europe is looking into sending data centres into space
- Published
- Jul 8 2024
- Word count
- 735 words
Ginormous caveat notwithstanding, don't data centers need a lot of cooling? How do you do that in space without massive radiators? Also, wouldn't their bandwidth be fixed at the time of launch preventing further improvements?
Sounds a lot like a "wouldn't it be cool if" idea that doesn't really make sense.
There is no other way. But moving lots of heat through radiation in space is a lot more practical than on earth. You've got a bigger temperature difference to work from and I suspect the heat rejection isn't the hardest part of this problem to solve.
If Microsoft can't make terrestrial underwater data centers make sense I'm extremely skeptical of space based data centers being close to practical in the near future. Sounds like a maintenance nightmare.
Space may be the worst place I can think of to build a data center. Every single one of a data centers needs are uniquely challenging in space. They need cooling, which you touched on, but cooling is way way easier on earth where radiators can use convection and conduction to dump heat. Using radiation only would mean you'd need massive radiators that face away from the space station.
Power is another huge one, data centers absolutely devour it. The ISS, the most complicated, expensive vehicle humanity has ever produced, and with the largest space based power generation capability we've ever produced generates about 200kw of power, which wouldn't even be enough for a small data center.
The third thing is network connections. You can't run fiber to space. You could maybe do something with lasers but that requires a ton of novel technology and complicated ground stations that don't exist.
Then there's maintenance. You have to do a multi million dollar space wall every time you want to replace a hard drive or swap an interface out.
There are dozens of other reasons why out of all the places to put a data center, space would probably be the worst one. I think it would make more sense to put a data center on the peak of Mt Everest honestly.
I'd be curious to hear the justification here and if there are any use cases they can think of that would make sense to do in space.
Don't forget about cosmic rays causing bit flips in memory! It's a huge issue!
Most computers in space are actually 2, 3 or 4 identical computers preforming the same calculation and comparing results. In addition to all the issues you brought up, a data center in space would be at least 1/3 as performant with the same hardware to unsure accuracy.
For most actual data center usage you’re right, it’s wholly inappropriate. But for coordination and communication with a whole network of space-based stations and robots I can see it making sense as a communications hub just to be closer to the action and not worry as much about attenuating signal with atmosphere. I’m imagining something like Thread/Matter for satellites and robots.
Of course, I don’t see meaningful need for any of that by 2036. And there’s a whole different set of engineering challenges involved with it. And I’m not even sure if just beaming all that compute down to Earth wouldn’t still be easier.
I agree it's absurd and it blows me away that the conclusion for the report projects 2035 as a reasonable estimate for when this could be feasible as if sorting out land and power on land isn't several orders of magnitude easier and can be done in the present.
Like it mentions that orbiting data centers could use solar which, certainly is a plus, but also not unique compared to what can be done on the ground unless they're assuming they'd lock it orbit so that it only faces the sun. Which again just adds complication.
Cooling the data center would absolutely be a challenge. I think it would make more sense to rely on refrigeration cycles to increase the temp for heat rejection which would let you get away with a smaller radiator, but still not a trivial system to design. However, I still stand by that being the easiest technical challenge faced. As you noted, maintenance for a space station sounds ridiculously expensive regardless of how you approach upgrades.
It makes more sense when the group writing the report (Thales Alenia Space) stands to make money milking the EU for more funding on this.
Also maintenance, hard drives fail constantly and need to be rotated out, and humans cannot live in microgravity for more than a couple months (and have long-term health problems as a result of doing so). How exactly are you performing replacements on failed drives?
and the ping would be worse than most any location on the ground
My knowledge comes mostly from binging projectrho,but as I recall cooling is a MASSIVE problem in space. The temperature difference might be big but there's also no medium for convection so you have to rely exclusively on radiation which is just not that quick.
Just realised I'm repeating what we already said. I guess my point is that this sounds unfeasible, radiators and coolant means more weight means a smaller payload etc.
It's tricky, but radiation heat transfer scales to the fourth power as a function of your source and sink temperatures, (T_source^4 - T_sink^4), so you do have that advantage in space compared to on earth. But your designs are much more limited than if you could also utilize convection to help move heat way since the only variables you really have to increase heat transfer is more radiator surface area or a higher source temperature. Which from a refrigeration stand point absolutely begins adding size, weight, and cost. So you're definitely right there. Just a logistics nightmare through and through.
It'd be a giant cosmic ray plinko machine.