- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
You must log in or register to comment.
What is up with this obsession with data centres in space?
I don’t know where it started. Maybe someone rich watched Altered Carbon or played CP2077. But now that the fantasy’s circling among billionaires, others want to cash in.
…From the firm’s perspective, if it’s hilariously impractical, so what? It’s Bezos’s money, and they’ll take it.
And they must know funding for space-based science (the actual practical application) is drying up.
You make a sun shield to block the scorching temperature of direct sun exposure, which will generate energy and keep the equipment in the chilling cold shade… although a huge investment, in the long run, it’s more practical than keeping them on land.
I’m not sure if this is satire (I mean no offense), but its exactly what I’m talking about.
There’s nothing ‘chilling’ about being in the shade in space. Radiating large amounts of power is immensely difficult and, in fact, a huge engineering challenge for speculative spacecraft designs, like reactor-powered rockets.
My favorite ‘speculative space travel’ site has a utterly fantastic writeup on this: https://www.projectrho.com/public_html/rocket/heatrad.php
As it points out, it’s already a costly challenge on the ISS and Space Shuttle:
…And that doesn’t even matter, because it’s utterly dwarfed by the stupendous cost of getting even a pound into LEO in the first place. Or the absolutely monumental cost of any kind of maintenance (or equipment writeoffs) if something breaks. Hence I’m ignoring other extreme engineering challenges, like bit-flips from cosmic radiation (which Nvidia tested on the ISS, and which gets worse as lithography gets smaller), or complexities of connecting structures in space (hence this firm).
https://www.projectrho.com/public_html/rocket/engines.php#rockettyranny
Hence, the only things that go in space are things that absolutely must, like low volume scientific equipment. Its more practical to just put servers under antarctica because it’s still at least 50X cheaper to install them, run them, and maintain them there.
Hence, my point. Folks like Bezos aren’t expected to know esoteric, theoretical stuff. They’re business folks. But they should know enough to ask someone qualified to asses their fantasies.
Every few decades technology severely affects the cost of everything.
edit: To people downvoting me, in the last decade the cost of sending stuff to space went from up to $20k/kg to $1.5k/kg.
Read Project Rho.
TL;DR: we’re discussing fundamental physics limitations here, like the rocket equation, heat flow, how radiation interacts with mass, things like that. Not technology challenges.
In sci fi that completely throws technological/engineering limitations out the window because everything’s designed by sun-sized transcendent AI, even they face the same limitations: https://www.orionsarm.com/eg-topic/49350e2d34113
I don’t have the technical knowledge to join the discussion, but wasn’t every technology we have today considered sci-fi at some point?
Is a huge heat shield with some aerogel or something behind it to contain the heat that couldn’t be turned into energy, and then a cord to transport the energy to the aerogel-coated equipment hundreds of meters away, really so unfeasible, as better aerogel and heat-to-energy conversion technology seem to be here?
Your second link seems to be about space travel. I’m just talking about having data centers orbiting Earth, like the thousands of satellites already do.
I will bet money that at some point someone argued that communications going through satellites in space would be unpractical because of engineering and technological challenges, the stupendous cost of getting stuff up there, and monumental maintenance costs compared to just having some lines going through some street poles.Let’s do math.
Let’s say the “space datacenter” peaks at 500 megawatts, seeing how Earth ones apparently peak around this. But that includes stuff like waste heat from power generation, the cooling and comms system, eveything on the spacecraft.
Lets say we want the coolant at 60C, so the computing stuff stays under 80C, as I am trying to give this system the benefit of the doubt. And lets assume the radiator is quite efficient and ignore mere engineering concerns, and give it an overall emissivity of 0.8.
From Project Rho:
Radiator area = P / (ε * σ * T^4)
Radiator area = (10 ^ 8 W) / ((0.8 emissivity) * (5.670374419 *10^-8, boltzman constant) * (333 Kelvin ^ 4, the same as 60C))
…That’s a radiator two thirds of a mile across.
Let’s, again, toss practicality out the window and say the “weight” of the whole thing is similar to a 6 mm aluminum panel, which seems like an unreasonable feat of engineering. After all, we gotta pump liquid through the thing, and unfold it somehow. But lets go with it.
That’s 5400 cubic meters of aluminum. That’s 5.6 * 10^6 kilograms. Picture a cargo ship flattened into a disk; that’s the order of mass we’re talking about.
At 20,000 kg per flight… that’s 112 Falcon Heavy flights to low earth orbit, or ~$10 billion dollars. Just to get our impossible radiator into orbit, and nothing else. Lets say launch costs get 10X cheaper, somehow, and that’s still a billion dollars.
Is a huge heat shield with some aerogel or something behind it to contain the heat that couldn’t be turned into energy
The first line of the Project Rho page:
NO, for the millionth time you CANNOT get rid of the heat by turning it into electricity!
And yes, we’re assuming the radiator is way far away on some kind of tether. Again, first few lines of the page. But the aerogel wouldn’t make a difference, it would actually hinder this system.
I will bet money that at some point someone argued that communications going through satellites in space would be unpractical because of engineering and technological challenges, the stupendous cost of getting stuff up there, and monumental maintenance costs compared to just having some lines going through some street poles.
No.
No one was saying this.
People complained about the engineering concerns, but engineers recognized it was physically practical… in theory.
But we are talking about physics now. Doesn’t matter if you’re a transcendental AI or not, you cannot engineer your way around thermodynamics, and we are talking about just one problematic system.
The only tech that would make such a dramatic difference (that’s on the horizon) is a space elevator, as it circumvents the rocket equation entirely by ‘pushing’ against the Earth. But this is really, really, really hard.
What hasn’t affected the cost of everything every few decades?
Given that one of the largest problems with the data centers we’re building today is heat dissipation, that seems like an exceptionally poor choice. Space creates major problems for heat dissipation.
The high radiation environment and the challenge of doing common maintenance tasks (e.g. disk replacement) seem prohibitively difficult as well…
Yeah. Honestly, I’m having a hard time thinking of any substantial benefits. Eventually, okay, sure, there’s a point in time where we can’t create computer structures on Earth if we’re going to scale up, but that is way the hell out there on the list of constraints we have. I also kind of suspect that materials science and manufacturing and computing technologies may change a lot and obsolete anything we create now long before that.
The article has:
“Starcloud’s mission is to move cloud computing closer to where data is generated,” Starcloud CEO Philip Johnston said in a statement.
But most data isn’t generated in space. It’s generated on Earth. Maybe if you have some kind of Earth-observation satellite in low earth orbit and want to add a shit-ton more processing capability to it so you don’t have to send its data back down to datacenters on Earth to chew on? Sounds kind of Orwellian, but maybe I could see that. But it seems like such a niche case.
You could put that on the dark side of the moon for example. Don’t get much colder than that. Radiating the heat away is no problem is you don’t take up much in the first place. The question is where energy comes from. I’m guessing they are going for nuclear energy for that. Sounds doable l, but why would you want to?!
The moon is tidally locked to the earth, not the sun. i.e., “the dark side” is the side that is never exposed to Earth, but it has a regular “day” cycle in the form of lunar phases. And dissipation would still be a problem, because you don’t have air to dump the heat that computers generate to
Radiative cooling. You can still have a data center in a stationary orbit between the earth and the moon thereby shielding it from the sun most of the time. But you are right, much harder problem. Gotta think on that some.
https://en.wikipedia.org/wiki/Kessler_syndrome
Maybe cold storage of huge amounts of data in geosynchronous orbit could be a not-terrible idea. But I guess they’ll want to keep latencies low and place them in LEO
Edit: Curiously, the last time I read this article several years ago it presented the consequence of making space completely inaccessible in the introduction (can’t remember if sourced or not), while now halfway though the article, under “Implications”, it says
The catastrophic scenarios predict an increase in the number of collisions per year, as opposed to a physically impassable barrier to space exploration that occurs in higher orbits. [1]
I wonder if “in 2025 the number [of tracked space debris] was estimated at over 11,800, most of which (7,135) belonged to Starlink” has anything to do with that 😒
citation needed ↩︎
