Want to get to Mars with a modern computer? It’ll be tough, because the radiation in space wreaks havoc with electronics.
Hewlett Packard Enterprise, though, has a new approach that’s showing promise as a way to cope with all those pesky protons blasting through our solar system.
Spacecraft designers for decades have “hardened” computers to withstand the onslaught of cosmic rays and other radiation in space. Unfortunately, hardening results in computers that run more slowly and take years to develop and test. The International Space Station’s command computers use Intel 386SX processors first released in 1985, for example, and hardened chips from aerospace giant BAE Systems run only at about a twentieth the clock speed of modern laptop processors.
It doesn’t necessarily have to be that way. Hewlett Packard Enterprise instead accepts that computers will be thrown for a loop, but then uses software to detect problems, assess their severity and more gracefully recover.
To test the approach, HPE pulled a standard server with dual Intel processors off its factory line and sent it up to the ISS for a year of testing. This machine, the Spaceborne Computer, for most of that test run.
“We’ve completed 345 days since in launch. It’s been working pretty well,” said Mark Fernandez, HPE’s Americas technology officer for high-performance computing.
It’s a curious modern link between sci-fi challenges and more down-to-earth computing concerns. The purpose of the Spaceborne Computer is to develop technology that will work on a lunar base or a mission to Mars. But the technology also should help computers running in the data centers that provide today’s biggest companies with computing power.
“We have seven potential patents evaluated now,” Fernandez said.
Computers that care for themselves
HPE calls its approach “autonomous self-care with fault detection and avoidance,” a useful if clunky name. With it, the computer monitors parameters like processor temperature and voltage, memory errors and the speeds of cooling fans. When the system finds trouble spots, it isolates them. And it dials back the performance to try to navigate through choppy waters.
“We had a mantra,” Fernandez said. “Running fast is better than running slow, running slow is better than running idle, running idle is better than being powered down, being powered down is better than being damaged.”
Another approach already in today’s server proved useful in space, too: HPE’s Memlog. This technology divides memory chips into countless small elements so that the computer can excise any given one it determines is headed for permanent failure. That nips bigger problems in the bud.
The machine has been running calculations around the clock, and it’s performed at a trillion mathematical calculations per second. After its year is up, NASA will ship it back down the gravity well to HPE engineers, who’ll scrutinize its innards to get detailed information on what exactly worked and what was damaged.
SSDs fail in space
One thing HPE learned: The solid-state disks (SSDs) used for data storage, which it expected to be among the most reliable components, turned out to be the weakest link.
“They’re failing at an alarming rate,” Fernandez said. “You wouldn’t want to take them to Mars without additional software protection like we did with other components.”
He hopes there will be a Spaceborne Computer 2. It could be used to test improvements the SSD makers think will make them as reliable as hoped, but the main goals would be getting a computer farther into space for a longer period of time.
It’s still a long way from a machine that’ll keep working all the way to Mars and back, but at least it’s not a system held back by the hardening process.
“This is the first baby step,” Fernandez said. “It shows we can take the iPhone X with us instead of the iPhone 2.”
: Sorry, Elon Musk.
: And new hope for (teeny tiny) life.