3 ms·
Surely the key issue that makes this untenable is network latency. That's a hard physical limit no matter how far our technology evolves to work around the myri
by seanhunter 6y ago
Surely the key issue that makes this untenable is network latency. That's a hard physical limit no matter how far our technology evolves to work around the myriad other problems with having servers in such an inconvenient location.
Maintenance is not a small problem though. Even if you were able to design servers to be robust for 40+ years, your users would have to be content with using 40+ year-old server tech as opposed to the latest and greatest. That's going to be ok for some use-cases but relatively few.
- ge96 6y ago> network latency I realize that's one of the first things that comes to mind. Either you beam power down or you beam compute down/"real time" communication... but I thought you could do something like "ML training" or anything else where the compute process does not need to be shared until it's done. On a side note, are the underwater servers better? I suppose that's easier to get to than space. The communication is there(transatlantic cables perhaps). Another side note, what if we could embed "servers" in an asteroid to hitch a ride... nvm the logistics of doing that. But another way to extend our existence/influence. RTG/solar for power.
- IAmEveryone 6y agoNo, the key issue is that launching servers and solar panels into space costs easily 10x as much as the hardware itself.