High-performance computing experiment to advance space travel

October 2nd, 2017, Published in Articles: EE Publishers, Articles: EngineerIT

Hewlett Packard Enterprise (HPE) and NASA have recently launched a supercomputer to the International Space Station (ISS) as part of a one-year experiment to test high-performance computing in space conditions.

The high-performance commercial off-the-shelf (COTS) computer system is called the “Spaceborne Computer”. It is unique in that it relies on software hardening, as opposed to the usual ruggedising of equipment to withstand space conditions including radiation, solar flares, subatomic particles, micrometeoroids, unstable electrical power, and irregular cooling. A software hardening approach saves time and reduces the cost and weight that result from hardware modifications.

This objective of the research is to verify whether the system can operate correctly by lowering its power and speed during high radiation events, and to help scientists identify ways of using software to protect ISS computers without protective shielding. This knowledge could help advance computing systems required for interplanetary travel, which will require powerful on-board processing to reduce the latency currently associated with sending and processing data on Earth.

Configuration and testing

The Spaceborne computer includes an HPE Apollo 40 class systems (from the company’s high-density server range) with a high-performance computing interconnect running an open-source Linux operating system. While it weighs approximately 56 kg on Earth, its weightlessness in space means tiny bolts can hold it in place. The supercomputer also has a water-cooled enclosure for the hardware.

Fig. 1: Spaceborne Computer, isometric view. (Credit: NASA)

Fig. 1: Spaceborne Computer, isometric view. (Credit: NASA)

Computing and data intensive applications will be run on the Spaceborne computer systems and its power consumption closely monitored and dynamically tuned during these runs. At the same time the effects of radiation on the systems will be determined with detecting, analysing and adapting to data, QuickPath Interconnect internal and FDR external errors.

The experiment will run four identical high-performance COTS computer systems – two aboard the ISS, and another two on Earth to act as the control group for the experiment. One of each pair will be run at a steady power/performance state while the second of each pair will be dynamically changed.

The control and monitoring will be done over a data network, for which HPE will use the Ku band internet protocol services and the approved Telescience Resource kit to monitor and control the two systems on the ISS.

The systems are expected to generate about 5,4 MB of uncompressed ASCII files per day, to be stored on-board the system’s internal solid-state disks and downloaded daily for safekeeping and analysis.

Fig. 2: Spaceborne Computer installed in Express rack, inverters not shown. (Credit: NASA)

Fig. 2: Spaceborne computer installed in Express rack, inverters not shown. (Credit: NASA)

The initial startup and setup has been successful according to HPE’s Mark Fernandez, the co-principal investigator of the experiment. In a blog post he describes how his team ran a multi-node High Performance Linpack benchmark test on the ISS systems, which has achieved one teraflop (i.e. one trillion calculations per second) performance.

Outcomes and expectations

This experiment will help identify critical failure points in electronic systems, as well as potential software patches that can prevent them. In addition to testing software hardening, the long duration experiment will also study the practicality of running and managing commercial off-the-shelf high-performance computer systems in orbit. Besides improving radiation-resistant computers for reliable computation in space, it can also help minimise risks from radiation such as solar flares in computing environments on Earth.

Acknowledgement

This article draws on writing by HPE and NASA, including:

[1] ‘Hewlett Packard Enterprise Sends Supercomputer into Space to Accelerate Mission to Mars’, HPE Newsroom. [Online]. Available: https://news.hpe.com/hewlett-packard-enterprise-sends-supercomputer-into-space-to-accelerate-mission-to-mars/. [Accessed: 01-Oct-2017].
[2] ‘HPE’s Spaceborne Computer Successfully Powers Up in Space and Achieves One TeraFLOP’, HPE Newsroom. [Online]. Available: https://news.hpe.com/hpes-spaceborne-computer-successfully-powers-up-in-space-and-achieves-one-teraflop/. [Accessed: 01-Oct-2017].
[3] ‘NASA – High Performance Commercial Off-The-Shelf (COTS) Computer System on the ISS’. [Online]. Available: https://www.nasa.gov/mission_pages/station/research/experiments/2304.html#overview. [Accessed: 01-Oct-2017].
[4] ‘One Small Step Toward Mars: One Giant Leap for Supercomputing’, HPE Newsroom. [Online]. Available: https://news.hpe.com/one-small-step-toward-mars-one-giant-leap-for-supercomputing/. [Accessed: 01-Oct-2017].

Send your comments to engineerit@ee.co.za

Subscribe to our leading email newsletters

FREE-OF-CHARGE

CLICK for other EE Publishers information products