First space station in LEO to match the ISS in mass?

[Robotbeat]

Because most electronics are not vacuum qualified.

Russians usually have used electronics in sealed nitrogen pressure vessels for this reason. The extra mass of the pressure vessel is not a problem compared to the overall costs of vacuum certification for stuff, especially with lower launch costs

 and being in vacuum would make servicing them harder. Air also helps cool components not directly liquid cooled.

[DanClemmensen]

It’s possible next generation large stations won’t be research-focused or tourism-focused or manufacturing-focused at all. They may be datacenters in pressurized volumes that have humans only for maintenance and upgrades (and may not be permanently inhabited… although if the datacenter is large enough…probably multiple Gigawatts…, this process would be continual and so you’d just keep people up there).

why bother with pressurized volume for that?  Just trusses with racks would be easier.

Yep. Why should I let messy stinky humans with their humid air into my nice clean vacuum-filled server room? If they need to come in here they can suit up.

[Robotbeat]

It’s possible next generation large stations won’t be research-focused or tourism-focused or manufacturing-focused at all. They may be datacenters in pressurized volumes that have humans only for maintenance and upgrades (and may not be permanently inhabited… although if the datacenter is large enough…probably multiple Gigawatts…, this process would be continual and so you’d just keep people up there).

why bother with pressurized volume for that?  Just trusses with racks would be easier.

Yep. Why should I let messy stinky humans with their humid air into my nice clean vacuum-filled server room? If they need to come in here they can suit up.

Does your laptop work in vacuum? Your GPU?

[Robotbeat]

Might make sense to have it nitrogen filled instead of air. But in either case, humidity can be completely scrubbed out of the atmosphere (edit: actually you might want some humidity for static electricity dissipation). Crews can use an oxygen mask. But vacuum makes design a headache.

[Robotbeat]

Relevant (but wrong scale for this)
https://www.datacenterdynamics.com/en/news/axiom-space-to-develop-orbital-data-center-on-its-commercial-space-station/

[Robotbeat]

Roughly a 20MW orbital data center would have the same mass as ISS’s 400t. Probably too small to be permanently crewed and maybe too small to be worthwhile, but there you go.

[Robotbeat]

I agree with those who think mass isn't a particularly good figure of merit. It's a decent proxy for how 'impressive' a station seems though. (Is being 'impressive' what ISS was all about?) The 'crew size' metric might be a good proxy for how useful a station is, assuming crew time has some utility.

The one I propose is 'mass flow rate' (though not exactly in the rocket engine sense). Combined up-mass and down-mass, maybe not just cargo but the mass of the visiting vehicles as well?

You could do something with how many person-months of occupancy does it support on orbit per year?

~Jon

I think I agree with this. Orbital data centers would basically just be man-tended satellites, mostly not used for habitation.

[Jim]

Because most electronics are not vacuum qualified.

Russians usually have used electronics in sealed nitrogen pressure vessels for this reason. The extra mass of the pressure vessel is not a problem compared to the overall costs of vacuum certification for stuff, especially with lower launch costs

 and being in vacuum would make servicing them harder. Air also helps cool components not directly liquid cooled.

they aren't going to use COTS server hardware, so just make vacuum rated

Being in the vacuum and not in can would be easier to service. No humans needed, just R&R the rack with the problem with a robotic spacecraft. And do the real repair at the spot on earth.

Cold plate cooling is easy

[Jim]

Does your laptop work in vacuum? Your GPU?

not that hard to fix.  just a larger heat sink/radiator

laptop form factor isn't good for a server anyways,

[Robotbeat]

Now you have to qualify all that stuff for outgassing, etc.

[Robotbeat]

Because most electronics are not vacuum qualified.

Russians usually have used electronics in sealed nitrogen pressure vessels for this reason. The extra mass of the pressure vessel is not a problem compared to the overall costs of vacuum certification for stuff, especially with lower launch costs

 and being in vacuum would make servicing them harder. Air also helps cool components not directly liquid cooled.

they aren't going to use COTS server hardware, so just make vacuum rated

using COTS hardware is the cheapest option. ESPECIALLY in the early days.

Being in the vacuum and not in can would be easier to service. No humans needed, just R&R the rack with the problem with a robotic spacecraft. And do the real repair at the spot on earth.

Cold plate cooling is easy

vacuum rated robotics are also much more expensive.

[jongoff]

Because most electronics are not vacuum qualified.

Russians usually have used electronics in sealed nitrogen pressure vessels for this reason. The extra mass of the pressure vessel is not a problem compared to the overall costs of vacuum certification for stuff, especially with lower launch costs

 and being in vacuum would make servicing them harder. Air also helps cool components not directly liquid cooled.

they aren't going to use COTS server hardware, so just make vacuum rated

using COTS hardware is the cheapest option. ESPECIALLY in the early days.

Being in the vacuum and not in can would be easier to service. No humans needed, just R&R the rack with the problem with a robotic spacecraft. And do the real repair at the spot on earth.

Cold plate cooling is easy

vacuum rated robotics are also much more expensive.

Using COTS hardware isn't just about it being the cheapest option. COTS electronics also tend to be significantly ahead of the performance of space rated hardware. It's also the only likely approach that easily scales. The amount of computing power in even a single 20MW data center would likely dwarf the total computing power that's been launched in space to-date. Maybe you could make something more space rated that had good performance, cost, and the ability to scale up to that quantity of processors fast enough. But frankly I think that putting them in a pressure vessel full of dry nitrogen seems a whole lot more likely to actually get you to an affordable and scalable space data center architecture.

(though to be fair, I'm still pretty skeptical about space data centers outside of edge computing myself, and for edge computing, I'd likely use something more like what Jim's talking about).

~Jon

[Jim]

Using COTS hardware isn't just about it being the cheapest option. COTS electronics also tend to be significantly ahead of the performance of space rated hardware. It's also the only likely approach that easily scales. The amount of computing power in even a single 20MW data center would likely dwarf the total computing power that's been launched in space to-date. Maybe you could make something more space rated that had good performance, cost, and the ability to scale up to that quantity of processors fast enough. But frankly I think that putting them in a pressure vessel full of dry nitrogen seems a whole lot more likely to actually get you to an affordable and scalable space data center architecture.

(though to be fair, I'm still pretty skeptical about space data centers outside of edge computing myself, and for edge computing, I'd likely use something more like what Jim's talking about).

~Jon

a pressure vessel with n2 in it is much different than a habitable module. 

But I am thinking the servers could be Starlink type electronics.    A server farm is going to be a lot of the same chips.  More than all that has flown to date (Starlink excluded).  If orbital server farms are viable, then developing space rated chips should be too. 

[Jim]

they aren't going to use COTS server hardware, so just make vacuum rated

using COTS hardware is the cheapest option. ESPECIALLY in the early days.
[/quote]hard

there are no COTS server pressure vessels.  Plenty of COTS spacecraft hardware to support vacuum rated servers.