HPE and NASA are working together to implement the first HPC system in space. Spaceborne Computer promises to accelerate HPC innovation and transform space exploration as we know it.
Since NASA was founded in 1958, the U.S. has made herculean achievements in human space flight, aeronautics, space science, and technologies that have revolutionized the way we think about and approach the universe. Today, history is still being made aboard the International Space Station (ISS).
The ISS serves as a space environment research laboratory in which crew members conduct and facilitate experiments in astronomy, life sciences, physics, and meteorology. The most recent resupply mission, that brought equipment and provisions to astronauts aboard the ISS, was launched on Monday, August 14th at 12:31 EDT at NASA’s Kennedy Space Center. The SpaceX CRS-12 cargo launch delivered 6,000 pounds of critical materials for more than 250 science and research investigations. Among these supplies was a high performance computing (HPC) system that promises to accelerate HPC innovation and transform space exploration as we know it.
LONG-RANGE SPACE TRAVEL
Hewlett Packard Enterprise (HPE) and NASA launched HPE Apollo 40 class systems, collectively known as the Spaceborne Computer, to implement the first fully functional HPC machine in space. The goal of this powerful collaboration is to provide interplanetary missions with the latest supercomputing capabilities that support simulation, artificial intelligence, and real-time data collection and analysis as well as reduce the downtime and expense of moving data back to distant Earth.
Sending electronic equipment to space is a time-consuming and costly process that involves physically hardening the technology against radiation and other environmental hazards. Because of this, compute systems may be multiple generations behind by launch time. Spaceborne Computer will instead experiment with commercial off-the-shelf (COTS) hardware that is hardened only by software to deliver HPC performance capabilities, far outpacing any human-built system currently in space.
HPE used two COTS HPC nodes loaded with advanced self-care software to oversee and protect their progress for a year-long experiment on the ISS. During this mission, Apollo servers will continuously run compute- and data-intensive HPC benchmarks in changing environmental conditions and monitor factors such as power consumption, while dynamically tuning energy consumed. Engineers will then compare the performance, runtime, and results of these machines with the output of two identical earth-based systems. This will allow HPE to determine the effects of harsh environmental factors like radiation on HPC machines and adapt them in real-time. The objective is to enable the launch of the latest COTS supercomputers—as is—for use on long-range space voyages.
DELIVERING THE FIRST SPACEBORNE COMPUTER
To house the HPC systems, Hardware Project Developer Dave Petersen and his team designed a metal locker that satisfies over 100 NASA safety and size requirements. The locker is powered by 110 volts of AC and 28 volts of DC from solar panels. Converters on the ISS convert 28VDC to 110VAC, using as many converters as necessary to reach the power envelope. Additionally, the locker features Triple-A Air (AAA) augmented with water-cooled heat exchangers that expel approximately three-quarters of its heat into the water—while the cooling loop extends into space, allowing the locker to operate like a hyper energy-efficient data center at zero dollars. Standard HPE Apollo 40 class servers were installed inside the locker which is now placed within the ISS EXPRESS Rack 6. This HPE locker can be used to continually refresh technology on board the ISS.
To achieve optimal performance and durability, Software Project Developer Mark Fernandez and his team developed a series of “umbrellas.” An umbrella loop sequences benchmarks in a particular order, time stamps them, and collects benchmark output data as well as system environmentals. When a loop is complete, a variety of data (i.e. environmental data, power data, error messages, etc.) is packaged into a bundle and transferred to Earth for post-analysis.
The loops are monitored by a second layer that ensures data movement occurs on an expected timeline. If the timeline falters, it alerts engineers on the ground to warnings and critical errors.
The third layer is a main loop which monitors the previous layers. The main loop can take action to continue an experiment or save hardware from severe damage. It has the ability to autonomously place servers in idle mode or a low-powered, low-heat generation mode to assess potential damages, while bundling data to transmit to Earth. In the event of environmental, power, or heat issues, the main loop will signal that it is shutting down payload and send the remainder of its data to HPE for critical analysis.
FROM EARTH TO THE NEXT FRONTIER OF SPACE EXPLORATION
Twenty years ago, the world saw its first one-teraflop machine that used 850 kilowatts of electricity—almost a megawatt. Two months ago, the first LINPACK-run teraflop computer using less than a kilowatt of free electricity and free cooling began operating in space.
Implementing a high performance spaceborne computer is a first for the HPC community, an effort that will enhance space exploration, fuel scientific research, and accelerate discoveries like never before. In the HPE continuum of computing solutions, from the Hybrid IT core to intelligent edge devices, the Spaceborne Computer is also the most remote edge server—that is, out of this world. What we learn from this year-long experiment may also be applicable to improve Earth-based computers and devices.
Don’t miss this revolutionary next step in space science. To learn more about the latest HPC innovations, I invite you to follow me on Twitter at @EngLimGoh. For ongoing news and updates on HPE solutions, you can also check out @HPE_HPC. And for up-to-the-minute information on Spaceborne Computer and the next frontier of space exploration, visit @NASA and @Space_Station.
Eng Lim Goh, PhD
VP, CTO – HPC & Artificial Intelligence
Hewlett Packard Enterprise