• Security incident: ISF was recently accessed by intruders. Please change your password, and change it anywhere else you used it. Read more

Cont: Musk, SpaceX and future of Tesla II

Now I'm no expert on any of this (although I was aware, from my compulsory [ie: childhood] science education, that a vacuum is not good for make hot go away). Having read the opinions of others that are, or are at least much closer than I am, I can only conclude that Noel is either stupider than me, or a liar, or both.

Again.

ETA: Regarding the highlighted, a delightful bit of autocorrect.
 
Last edited:
There are so many issues, one is certainly cooling, but another is interconnection. I can see how the delay from earth to satellite may not be a huge issue, you don't use the space cloud for stuff that needs "instantaneous" responses but that still leaves the problem of connecting the small data centres together.

And as for the cooling yes there are ways to do some cooling in space but the server racks necessary on the earth are stretching what cooling can cope with and those options are simply not possible in space.

Plus of course sooner or later we will achieve the density of satellites to see the Kessler syndrome in action.
 
Darat said:
There are so many issues, one is certainly cooling, but another is interconnection. I can see how the delay from earth to satellite may not be a huge issue, you don't use the space cloud for stuff that needs "instantaneous" responses but that still leaves the problem of connecting the small data centres together.

And as for the cooling yes there are ways to do some cooling in space but the server racks necessary on the earth are stretching what cooling can cope with and those options are simply not possible in space.

Plus of course sooner or later we will achieve the density of satellites to see the Kessler syndrome in action.
Click to expand...
Racks of servers are very good heaters. Very expensive heaters.
 
Darat said:
There are so many issues, one is certainly cooling, but another is interconnection. I can see how the delay from earth to satellite may not be a huge issue, you don't use the space cloud for stuff that needs "instantaneous" responses but that still leaves the problem of connecting the small data centres together.
Click to expand...
This is pretty much why Microsoft pulled the plug on their Burger King on the ocean floor pilot, if I remember correctly. Underwater data centers are technically feasible, but aren't scalable enough to be practical.

Orbital data centers see those results and say "We can do something about the technically feasible part."
 
self replicating optimus robots on spacex launches all funded by taxpayer dollars of course. he’d make trillions if it wasn’t stupid nonsense from a con man.

this is h2o gas for your car level stuff imo
 
Darat said:
There are so many issues, one is certainly cooling, but another is interconnection. I can see how the delay from earth to satellite may not be a huge issue, you don't use the space cloud for stuff that needs "instantaneous" responses but that still leaves the problem of connecting the small data centres together.
Click to expand...
AI and other high-performance, large-scale computation absolutely requires near-instantaneous, error-free, high-capacity communication among its computing nodes in order to achieve performance goals. When we build a supercomputing cluster, we look at achieving symmetrical bisectional bandwidth. This means that for a set of interconnected computers we can partition the set into two subsets and assure that each computer in one subset can communicate at a certain speed with each computer in the other subset. Then do that for all possible partitionings. We use products like Infiniband to engineer this. So that application is right out. All you get with a satellite-based topology is AWS in Space.
 
Darat said:
And of course you have the added problem of upgrading the servers in space. Perhaps we send up Optimus robots to do the maintenance work...
Click to expand...
For AI and other high-performance computing, this problem becomes acute. If you start out with N nodes, all newly manufactured, each node has a certain MTBF dictated by node model. This is a real phenomenon. When you have thousands of nodes whose MTBF clock starts at the same time, you can actually see the failure rate increase toward the mean in the first cycle. The probability density function for node failure is the infamous Bathtub Curve.

In distributed high-performance computing, all the nodes that participate in the computation must remain functional. The failure of a single node can stall the computation until the node is replaced, either by bringing up a standby node or by replacing the node physically. Either way you have to start the computation over from the last checkpoint.

As you replace nodes whenever they fail, the MTBF clocks for new nodes start over at increasingly distributed times. MTBF for the entire cluster grows shorter. In one case years ago, we had a supercomputing cluster with a mean uptime of only 7 hours. The MTBF clocks become too jittered. At a certain point you simply have to replace all the nodes with newly manufactured ones, regardless of their health.
 
mumblethrax said:
This is pretty much why Microsoft pulled the plug on their Burger King on the ocean floor pilot, if I remember correctly. Underwater data centers are technically feasible, but aren't scalable enough to be practical.
Click to expand...
On the other hand, the ocean is a large thermal mass that is available to dump heat into.
 
arthwollipot said:
On the other hand, the ocean is a large thermal mass that is available to dump heat into.
Click to expand...
Yeah, it's a much more appealing idea on its face, but it faces some of the same drawbacks (like inability to service equipment one it's deployed). China has at least one commercial underwater data center, but it's still pretty new.
 
mumblethrax said:
Yeah, it's a much more appealing idea on its face, but it faces some of the same drawbacks (like inability to service equipment one it's deployed). China has at least one commercial underwater data center, but it's still pretty new.
Click to expand...
Build the data centre on a beach and use pumped sea water for cooling. That seems a little cheaper than putting the whole thing under water.
 
jeremyp said:
Build the data centre on a beach and use pumped sea water for cooling. That seems a little cheaper than putting the whole thing under water.
Click to expand...

probably don’t want to use sea water due to the corrosion from the salt.
 
dirtywick said:
probably don’t want to use sea water due to the corrosion from the salt.
Click to expand...
Heat exchangers.

Quite a few nuclear power stations in the UK are on the coast to take advantage of the cold sea water. You will notice a lack of cooling towers

Old and new
Hartlepool nuclear power station with the Tees Bay wind farm in the distance.

I could see the Hartlepool station across Tees Bay from my bedroom window as a kid

1774531383011.jpeg
 
GraculusTheGreenBird said:
Yep, not like we need to worry about any side effects from that.. ;)
...
Click to expand...
You mean like changes to the local flora and fauna caused by the warmer water. Including algae and such, which could clog up...
dirtywick said:
probably don’t want to use sea water due to the corrosion from the salt.
Click to expand...
some sort of clever heat-exchanger sytem?


Oops! Semi-ninjaed by Andy
 
Last edited:
Andy_Ross said:
Heat exchangers.

Quite a few nuclear power stations in the UK are on the coast to take advantage of the cold sea water. You will notice a lack of cooling towers

Old and new
Hartlepool nuclear power station with the Tees Bay wind farm in the distance.

I could see the Hartlepool station across Tees Bay from my bedroom window as a kid

View attachment 69903
Click to expand...

oh yeah idk much about it. i just saw pump and sea water and thought pumping sea water is pretty hard on pumps and piping
 
You must log in or register to reply here.

Share this page

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom