NASA teams up with HPE to send a supercomputer into space ahead of a mission to Mars
Yesterday, SpaceX launched a Dragon Spacecraft on a resupply mission to the International Space Station from Cape Canaveral. On the surface of things, this is about as close to mundane as space travel gets, but for a team of Hewlett Packard Enterprise engineers, it has special significance. On board was an HPE super-computer – appropriately enough from the Apollo 4000 range – especially kitted out for the journey ahead.
What’s impressive is that, despite NASA’s stringent safety requirements (the computer passed at least 146 tests to be certified by the space agency), the core technology hasn’t actually changed that much from what you could buy here on Earth, assuming your pockets are deep enough. Generally speaking, computers need to be ruggedised to deal with the unique problems space presents a standard computer with (radiation, solar flares, subatomic particles, micrometeoroids, unstable electrical power, irregular cooling and all that jazz), but HP took a different approach: the Linux-based software will throttle the computer based on real-time conditions, ensuring the computer can – in lab conditions at least – dodge errors brought on by its unusual office conditions.
“The data downlink from the ISS is a challenge,” Dr Mark Fernandez, Americas CTO of SGI at HPE, told Alphr. “It’s not as fast as we would like and there are areas of LOS [loss of signal]. When the ISS is in an LOS condition, we detect that and hold the data onboard until signal is restored. This was an unexpected hurdle, but like the other aspects of hardening with software, we used software to overcome this network anomaly.”
There are two big positives from this new experiment, assuming it all goes to plan. First, traditional ruggedisation takes time and means that computers are often older and slower than current Earth-bound models by the time they go up. “One of the goals of this experiment to see if we can send to space a current-generation system and harden it with software,” says Fernandez. “This is much faster and less expensive, and results in current-generation computing power in the hands of space explorers and scientists.”
Second, ruggedisation adds unnecessary weight: not a problem when you reach space, but anything that reduces the size of the payload before liftoff is certainly welcome. Through this approach, HPE’s precious cargo weighs just 124 pounds, and still matches NASA’s requirements – on Earth, at least. “We were provided with a maximum electrical power available, air-cooling capability and chilled water characteristics,” explains Fernandez. “No compromises were made per se. Like all HPC configurations, we of course would like more power, more cooling and more space. But we will be able to meet the objectives.”
The system is water-cooled – and the water is actually provided by NASA, and connected to the chilled water loop onboard the ISS, which eventually extends out to space. “It’s going to be the most efficient computer ever, because A) it’s powered by solar cells, with zero cost of electricity, and B) it’s cooled by the coldness of space at zero dollars as well,” explains Fernandez.
“If someone tried to do the energy-efficient computations, they’d have to divide by zero somewhere.”
“It’s going to be the most efficient computer ever, because A) it’s powered by solar cells, with zero cost of electricity, and B) it’s cooled by the coldness of space at zero dollars as well”
Make no mistake: this could be hugely important for the future of space travel. Right now, astronauts are limited in terms of the computing power they have access to in space – in fact, many calculations are still done on Earth. To date, this hasn’t really been a problem because the moon and International Space Station are – galactically speaking – right on our doorstep. When our astronauts get further away from Earth, delays in communications could cause real problems. With astronauts exploring Mars, we could be talking about a 20-minute delay between communications reaching Earth, and another 20 before the response returns to our explorers. An awful lot can go wrong in 40 minutes, and it’s a time delay we could do without.
That isn’t to say that the benefits of a successful trial won’t be helpful with our less daring space missions. Will we see improvements to experiments on the ISS, or other missions closer to home? “Absolutely, that is a possibility,” enthuses Fernandez. “Like times before when a supercomputer has been made available to scientists, engineers and researchers, the results can be something no-one expected. There’s no need to ‘wait’ for a flight to Mars to potentially begin exploiting HPC capabilities aboard the ISS.”
For now, though, just making sure the computer works fine is the priority. “This project is in itself an experiment, so we’ll be eagerly awaiting the results of how it performs in the challenging conditions of space, and reviewing the entire project in real depth on its return,” concludes Fernandez.