HPE's Foray into Space with Spaceborne Computer

[DooKey]

HPE is currently working with NASA to launch a supercomputer into orbit to see how long it lasts so it can be used in deep space travels. Basically, they want to use the massive compute capabilities to monitor a number of data points, run applications, and also run analytics to figure out issues on board the spacecraft. The reasoning behind this is to try to fight events in space and not have to rely on ground based computers to do so. I'm actually surprised this hasn't already been done and I can't wait to see how it works out.

As space missions get more advanced, there is a growing need for sophisticated onboard computing resources capable of functioning in the harsh conditions of space. By achieving one teraFLOP in space, HPE’s Spaceborne Computer has shown that astronauts now have access to high performance computing systems that can help them solve any mission critical scenarios in space.

 

[Elf_Boy]

My guess is it will crash like a 15 year old with two bottles of vodka and an expensive car on a joyride.

If they built a really good Faraday cage maybe only 1 bottle.

 

[Elf_Boy]

How are they going to power a supercomputer in orbit? Or dissipate heat?

 

[CombatChrisNC]

How are they going to power a supercomputer in orbit? Or dissipate heat?

Well, solar panels are really efficient up there compared to down here under all this air.

And the side of the craft not facing the sun would be very cold and a wonderful place to radiate any extra heat.

 

[Biznatch]

My guess is it will crash like a 15 year old with two bottles of vodka and an expensive car on a joyride.

If they built a really good Faraday cage maybe only 1 bottle.

How are they going to power a supercomputer in orbit? Or dissipate heat?

Faraday cage is to block radio signals. It's the radiation that will damage the computer and is what needs to be shielded.

And it will be floating in sub zero temps with no air. They'll cool it just like the computers on the ISS or any other device orbiting the earth with electronics.

 

[sfsuphysics]

Well radiation transfer goes roughly as T^4 if you radiate to space (effectively 0 temp, subtract 3^4 if you're really that anal). So depending on how hot it gets, room temp to space is roughly 450 watts per square meter which doesn't sound like a lot for a super computer, but if you go upwards of 400F that's nearly 3kW per square meter it'll radiate.

Presumably they'll have some liquid cooling that radiates the energy, and possibly goes through a compression cycle like a refrigerator so it can cool quicker.

 

[Poseur]

HPE is Hewlett-Packard Enterprise, for those of us that didn't know. Too many mergers and acquisitions and renaming companies.

 

[lcpiper]

In for a penny, in for a pound.

If one is going to launch a supercomputer into space to provide computing as a service to space borne consumers, the maintenance and upgrade elements need to be figured out. And as long as SpaceX is making so many flights all the time, I recommend they take a distributed computing approach and instead of launching one gigantic supercomputer, that they use unallocated cargo space in scheduled SpaceX flights to launch dozens or even hundreds of small computers that will network and share workload. As they age and become too dated, just replace them a few at a time rotating in the new while you deactivate the oldest.

Now a big machine has the benefit of being a single housed solution and for power you could give it a nuclear reactor if you wanted to, but like all things that go up, sooner or later it has to come back down. But many smaller machines could be powered via solar, wireless charging, etc.

Let's say it takes a device four days to charge, for one day of use. No problem, send up 500 devices and they take turns with work cycles and charging cycles.

All they have to do is start from the concept of adapting cell phones as distributed computing devices and work out the kinks from there.

I have an ulterior motive, I want more money for my old cellphones

 

[XenIneX]

Right... "Supercomputer." Way to oversell it...

You can get a teraFLOP out of a single high-end Xeon. A fully-redundant cluster could easily fit in a single 2u chassis.


If I can pick it up, it's not a supercomputer.

 

[ChadD]

If I can pick it up, it's not a supercomputer.

Would the "supercomputer" moniker not depend a bit on when its being used. In 1985 Crays best could push 2.4 gflops... in 96 Hitachi broke 300... in 97 Intel broke 1000... in 99 intel broke 2000... in 2004 IBM hit 70 Tflops... in 2007 IBM hit 478 TFlops. And of course today China is destorying everything with Sunways 93 PFlops.

Still for off the rack stuff today one xeon phi can easily break 900 gflops. HP and Nasa are only aiming for a terraflop of performance... so although I agree that is pretty hard to classify as a modern "supercomputer" it would put it on par with Intels room size "super computers" from the late 90s-early 00s. I'm sure that should be more then enough for them to calculate trajectories and the like should they somehow loose all communication ect.

 

[XenIneX]

Would the "supercomputer" moniker not depend a bit on when its being used. In 1985 Crays best could push 2.4 gflops... in 96 Hitachi broke 300... in 97 Intel broke 1000... in 99 intel broke 2000... in 2004 IBM hit 70 Tflops... in 2007 IBM hit 478 TFlops. And of course today China is destorying everything with Sunways 93 PFlops.

Still for off the rack stuff today one xeon phi can easily break 900 gflops. HP and Nasa are only aiming for a terraflop of performance... so although I agree that is pretty hard to classify as a modern "supercomputer" it would put it on par with Intels room size "super computers" from the late 90s-early 00s. I'm sure that should be more then enough for them to calculate trajectories and the like should they somehow loose all communication ect.

My watch has as much computing power as a ~1980 supercomputer. That does not mean I have a supercomputer on my wrist.

We know from HPE's marketing puff piece and NASA's mission brief exactly what they sent up there. It's two off-the-rack dual-socket Broadwell servers in a funny chassis. They're not even a cluster; just two separate machines configured a bit differently to test radiation-mitigation strategies.

That's not a supercomputer. It's just a computer.

 

[CombatChrisNC]

We know from HPE's marketing puff piece and NASA's mission brief exactly what they sent up there. It's two off-the-rack dual-socket Broadwell servers in a funny chassis. They're not even a cluster; just two separate machines configured a bit differently to test radiation-mitigation strategies.

That's not a supercomputer. It's just a computer.

Well, doesn't it make sense to test with something small before moving to something larger?

And compared with the laptops that they have, it's a hell of a lot more computing power than anything else up there.