- HPE’s Spaceborne Computer-2 will be flown to the ISS on February 20
- It will replace Spaceborne Computer-1, a machine which was meant to last one year but lasted nearly two
- Spaceborne 2’s machine learning and edge computing capabilities allow for a wider range of possibilities for the computer’s use
With all our Earthly pursuits, it’s easy to forget that a multi-billion-dollar spacecraft is constantly suspended in low-Earth orbit above our heads — or below our feet, depending on where the International Space Station (ISS) is during its roughly 90-minute journey around the planet.
But Dr. Mark Fernandez, principal investigator for the Spaceborne Computer-2 project at Hewlett Packard Enterprise (HPE), has been thinking about it for years. He believes that if scientists want to get serious about understanding the cosmos, they’ll need some serious computing power.
As such, he and his colleagues are the scientists behind space’s first supercomputer — Spaceborne Computer-1, which was conceived as an experiment.
“Spaceborne Computer-1 was ‘can you do it at all?’ It was a proof-of-concept,” says Fernandez. They wanted to see if they could make a supercomputer that could not only survive the harshness of space, but also make it through the treacherous ride up to the ISS.
After the mission’s success, the scientists embarked on their next step: readying Spaceborne Supercomputer-2 for launch on February 20, 2021. “Now, we’re wondering, ‘ok, you can do it, but can you do it for a Mars mission which is two to three years long?'” says Fernandez.
With Spaceborne Computer-2’s machine learning and edge computing capabilities, combined with the addition of graphics processing units (GPUs), such a goal isn’t just possible – it's within our reach.
Out with the old
Much like its predecessor, Spaceborne Computer-2 isn’t built with unique parts. In fact, part of the mission was building a robust machine with off-the-shelf devices.
“I can tell you where you can go purchase the identical servers that are in Spaceborne Computer-2,” says Fernandez. “That's how not special they are.”
What’s more, the first mission was so successful that a second machine seems to have been an inevitability. Fernandez states that Spaceborne Computer-1 returned to Earth in June of 2019 and a contract was signed for Spaceborne Computer-2 in October of 2019.
Spaceborne Computer-1 was originally meant to be a one-year mission, but the success of the machine allowed HPE to run it for nearly two years. Now that we know this device works as intended, we can begin to move our eyes to the next prize.
“When there's real rocket scientists or real astronauts in the room, I don't say the following: Most of our missions have been to get there and get back,” says Fernandez. “All the missions to Mars, we collected some rocks and brought them back. We didn't really do anything there.”
Fernandez continues: “Well, we're not going to take a year to get to Mars, land, and then turnaround and come right back. We've got to live there – and the crew is going to have to be more than pilots, more than astronauts. So, I call them space explorers.”
We’ve pointed out before that Fernandez uses the term “space explorers” rather than “astronauts,” and it seems to underscore the majesty of the Spaceborne Computer project. This isn’t about slapping a machine in a low-Earth-orbit satellite and calling it a day. This is about becoming an interplanetary species, and the tools we’ll need to get there.
Computing on the edge
Of course, many of the tools we’ll need are going to come in the form of computing – specifically, machine learning, artificial intelligence, and edge computing.
“With Spaceborne 1, edge computing and artificial intelligence and machine learning were in their infancy – now they're commonplace,” says Fernandez. “That’s why we've incorporated this technology into Spaceborne Computer-2. They asked us to send up 2x the capability.”
Fernandez continues: “We’ve got the edge line system on board, and inside of it is a GPU. In 2017, a GPU was hundreds of watts of power. Now, I'm putting in a 70-watt GPU. We'll be able to do general-purpose enterprise class computing on Spaceborne Computer-2 as well as GPU-enabled AI and machine learning, image processing, signal processing, et cetera.”
Although all of these technologies are being used in interesting ways, Spaceborne Computer-2’s edge computing capabilities are especially exciting. To make this point, Fernandez references a scene in The Martian. At one point, Matt Damon’s character has sent a message to Earth. It takes 24 minutes for the message to reach home, and then another 24 minutes to get back to him.
“About an hour later, he jumps and he's excited that he's in communication,” recounts Fernandez. “Well, in the movie, a big dust storm comes in and wipes out his whole farm. But suppose he were able to do the Mars weather predictions himself. That could be life-changing, or even life-saving. Self-sufficient computers enable self-sufficient explorers.”
And there’s that word again — explorers.
Explorers are many things — courageous, smart, capable — but their most important feature is curiosity. This drive to learn more and see more is at the heart of exploration, and Fernandez hopes this project will help our smartest people to better understand the universe we live in.
As of the writing of this article, HPE is putting out a call for submissions for research considerations on Spaceborne Computer-2.
“We will be providing Spaceborne Computer-2 as a service, and that's a new model,” says Fernandez. “Individual payload developers can’t afford to send up their own computational edge platform. But those that have the knowledge to say, ‘I can process this ISS data faster with this container than with someone else’s container’ – they can participate. It’s really a low-cost entry for the masses, for humanity, to help advance space exploration.”