23
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 01 Sep 2024
23 points (100.0% liked)
TechTakes
1485 readers
124 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
I read the white paper for this data centers in orbit shit https://archive.ph/BS2Xy and the only mentions of maintenance seem to be "we're gonna make 'em more reliable" and "they should be easy to replace because we gonna make 'em modular"
This isn't a white paper, it's scribbles on a napkin
there’s so much wrong with this entire concept, but for some reason my brain keeps getting stuck on (and I might be showing my entire physics ass here so correct me if I’m wrong): isn’t it surprisingly hard to sink heat in space because convection doesn’t work like it does in an atmosphere and sometimes half of your orbital object will be exposed to incredibly intense sunlight? the whitepaper keeps acting like cooling all this computing shit will be easier in orbit and I feel like that’s very much not the case
also, returning to a topic I can speak more confidently on: the fuck are they gonna do for a network backbone for these orbital hyperscale data centers? mesh networking with the implicit Kessler syndrome constellation of 1000 starlink-like satellites that’ll come with every deployment? two way laser comms with a ground station? both those things seem way too unreliable, low-bandwidth, and latency-prone to make a network backbone worth a damn. maybe they’ll just run fiber up there? you know, just run some fiber between your satellites in orbit and then drop a run onto the earth.
everyone who's ever done physical cabling knows aaallll about dropping cables upward
Easy, the cables go into the space elevator. Why do you all have to be so negative, don't you have any vision for the future?
what if my vision for the future is zeppelin data centers constantly hovering over the ocean? they’ll have to be modular, of course, and we can scale our deployment by just parallel parking a new zeppelin next to our existing one and using grappling hooks and cargo straps to attach the zeppelins to each other. as you can clearly see, this will allow for exponential growth! and networking is as simple as Ethernet between the zeppelins and dropping an ocean-grade fiber cable off the first zeppelin and splicing that into an intercontinental backbone link. so much more practical than that orbiting data centers idea!
ez
You're entirely right. Any sort of computation in space needs to be fluid-cooled or very sedate. Like, inside the ISS, think of the laptops as actively cooled by the central air system, with the local fan and heatsink merely connecting the laptop to air. Also, they're shielded by the "skin" of the station, which you'd think is a given, but many spacebros think about unshielded electronics hanging out in the aether like it's a nude beach or something.
I'd imagine that a serious datacenter in space would need to concentrate heat into some sort of battery rather than trying to radiate it off into space. Keep it in one spot, compress it with heat pumps, and extract another round of work from the heat differential. Maybe do it all again until the differential is small enough to safely radiate.
while radiating out waste heat at higher temp would be easier it'll also take up valuable power, and either i don't get something or you're trying to break laws of thermodynamics
I'm saying that we shouldn't radiate if it would be expensive. It's not easy to force the heat out to the radiators; normally radiation only works because the radiator is more conductive than the rest of the system, and so it tends to pull heat from other components.
We can set up massive convection currents in datacenters on Earth, using air as a fluid. I live in Oregon, where we have a high desert region which enables the following pattern: pull in cold dry air, add water to cool it further and make it more conductive, let it fall into cold rows and rise out of hot rows, condition again to recover water and energy, and exhaust back out to the desert. Apple and Meta have these in Prineville and Google has a campus in The Dalles. If you do the same thing in space, then you end up with a section of looped pipe that has fairly hot convective fluid inside. What to do with it?
I'm merely suggesting that we can reuse that concentrated heat, at reduced efficiency (not breaking thermodynamics), rather than spending extra effort pumping it outside. NASA mentions fluid loops in this catalogue of cooling options for cubesats and I can explain exactly what I mean with Figure 7.13. Note the blue-green transition from "heat" to "heat exchanger"; that's a differential, and at the sorts of power requirements that a datacenter has, it may well be a significant amount of usable entropy.
okay so you want to put bottoming cycle thermal powerplant on waste heat? am i getting that right?
so now some of that heat is downgraded to lower temperature waste heat, which means you need bigger radiator. you get some extra power, but it'd be a miracle if it's anything over 20%. also you need to carry big heat engine up there, and all the time you still have to disperse the same power because it gets put back into the same server racks. this is all conditional on how cold can you keep condenser, but it's pointless for a different reason
you're not limited by input power (that much), you're more limited by launch mass and for kilogram more solar panels will get you more power than heat engine + extra radiators. also this introduces lots of moving parts because it'd be stirling engine or something like that. also all that expensive silicon runs hot because otherwise you get dogshit efficiency, and that's probably not extra optimal for reliability. also you can probably get away with moving heat around with heat pipes, no moving parts involved
also you lost me there:
okay this works because water evaporates, cooling down air. this is what every cooling tower does
no it doesn't (but it doesn't actually matter)
and here you lost me. i don't think you can recover water from there at all, and i don't understand where temperature difference comes from. even if there's any, it'd be tiny and amount of energy recoverable would be purely ornamental. if i get it right, it's just hot wet air being dumped outside, unless somehow server room runs at temperatures below ambient
also i'm pretty sure that's not how it works at all, where did you get it from
and I’m over here like “what if we just included a peltier element… but bigger” and then the satellite comes out covered in noctua fans and RGB light strips
I was also momentarily nerdsniped earlier by looking up the capacity of space power tech[0] (panel yields, battery technology, power density references), but bailed early because it'll actually need some proper spelunking. doubly so because I'm not even nearly an expert on space shit
in case anyone else wants to go dig through that, the idea: for compute you need power (duh). to have power you need to have a source of energy (duh). and for orbitals, you're either going to be doing loops around the planetoid of your choice, or geostationery. given that you're playing balancing jenga between at minimum weight, compute capacity, and solar yield, you're probably going to end up with a design that preferences high-velocity orbitals that have a minimal amount of time in planetoid shadow, which to me implies high chargerate, extremely high cycle count ceiling (supercaps over batteries?), and whatever compute you can make fit and fly on that. combined with whatever the hell you need to do to fit your supposed computational models/delivery in that
this is probably worth a really long essay, because which type of computing your supposed flying spacerack handles is going to be extremely selected by the above constraints. if you could even make your magical spacechip fucking exist in the first place, which is a whole other goddamn problem
[0] - https://www.nasa.gov/smallsat-institute/sst-soa/power-subsystems/ (warning: this can make hours of your day disappear)
dusk-dawn orbit is a thing if you don't care too hard about where exactly to put it
but it's gonna be so fucking expensive, what they're trading off so it's even remotely worth it? do they think it's outside of any jurisdiction?
yeah I thought about that but I took it in light of "data center", i.e. presuming that you'd want continuous availability of that. part of what I mean with it being worth a long essay - there's a couple of ways to configure the hypothetical way this would operate, and each has significant impacts on the shape of the thing
yep. that's the thing that's so wild about this fairy picture. option 1) make your entire compute infra earthside[0], launch it all, and get .... the node compute equivalent of 3 stacked raspberry and a 2017 gpu, at a costpoint in the high 4 digits or more... or option 2, where you just shove a dc full of equipment for the price of like 20 such nodes, and have the compute equivalent of a significant number of mid-range hosters
even if (and this is extreme wand waving) you could crack non-planetbound production for the entire process and fab all this shit in space (incl. the mining and refining and ....) as a way to reduce costs, you still have all these other problems too. and it's not like this is likely to happen any time soon
guess they better hope 'ole ray has another vision soon, to get a fixed date for the singularity. can't see how you do your scrum planning for this fantasy without a target date provided by the singularitian prophet
wait it's all ray kurzweil?
dunno if the aforementioned jazz is (I didn't check), but rayboi is the easiest "and then compute things just become magically solved" touchstone for me to remember
too many of the fucking nutjobs to properly track who's the steering committee for each insane idea
BasicSteps™ for making cake:
Any further details are self-evident really.
every flat, sad looking chocolate cake I've made
Design principles for a time machine
Yes, a real, proper time machine like in sci-fi movies. Yea I know how to build it, as this design principles document will demonstrate. Remember to credit me for my pioneering ideas when you build it, ok?
You joke, but my startup is actually moving forward on this concept. We already made a prototype time travel machine which while only being able to travel forward does so at a promising stable speed (1). The advances we made have been described by the people on our team with theoretical degrees in physics as simply astonishing, and awe-inspiring. We are now in an attempt to raise money in a series B financing round, and our IPO is looking to be record breaking. Leave the past behind and look forward to the future, invest in our timetravel company xButterfly.
Fuck it, throw some more junk into orbit, why not
Who knew that the VC industry and AI would produce the most boring science fiction worldbuilding we will ever see
Who is even asking for this?