Technology is rapidly transforming the way we think and understand the universe. In the realm of space science, technological advancements have enabled man to study Earth, explore new worlds, and even walk on the moon. Today, high performance computing (HPC) is accelerating space research and scientific discovery like never before. Organizations like NASA are using the most cutting-edge HPC solutions to analyze troves of complex data, conduct experiments, advance the Mission to Mars, and much more.
Exploring new frontiers with HPC
NASA has teamed up with leading manufacturer Hewlett Packard Enterprise (HPE) to bring the most powerful earth-based systems into orbit in a mission called Spaceborne Computer. This year-long experiment is designed to test the durability and performance of high performance commercial off-the-shelf (COTS) computer systems on board the International Space Station (ISS). Scientists are utilizing these two HPC machines to run compute-intensive and data-heavy applications, monitor power consumption, and study the effects of radiation while running performance applications. A dashboard records benchmarks and other statistics—such as how many iterations are running and how many have been completed—and transmits the data back to Earth where HPE compares the results with two identical COTS systems. The data is then visualized by the ISS Track Viewer, a piece of software that updates every five seconds to monitor ISS movements and predict its potential track.
In preparation for space travel, HPE created specialized software to monitor and protect the hardware against radiation and other potential hazards. Instead of a physical hardening process, HPE software “hardens” the computer digitally to dramatically reduce time-to-launch and keep the hardware from becoming obsolete. The systems were launched to the ISS on August 14th, 2017 on the SpaceX Dragon spacecraft. After being without electricity and exposed to extreme cold and radiation for an entire month, the machines were installed and powered up on September 14th.
After 75 days and over 1,500 revolutions around Earth, the systems have detected two anomalies and one power failure. Designed in an active/active configuration, the remaining supply assumed the full electrical load, preventing any downtime since the experiment began. Based on this proven redundancy, NASA is currently working with HPE to launch a spare power supply to the ISS in 2018. HPE also has plans to begin shipping up space replaceable units (SRUs).
Today, both machines have maintained power and are running three benchmarks:
- High Performance LINPACK (HPL) – HPL is a portable software package that measures the floating-point rate of execution of distributed-memory computers by solving a dense system of linear equations.
- High Performance Conjugate Gradient (HPCG) – HPCG is designed to access computational and data patterns that more closely match a diverse set of critical applications. This benchmark is currently used to rank the TOP500 computing systems.
- NASA Advanced Supercomputing Parallel Benchmarks (NPB) – NPB are a set of benchmarks that help researchers evaluate the performance of parallel supercomputers.
Redefining HPC capabilities
Successful space exploration beyond Earth’s orbit will require sophisticated HPC solutions that can operate for extended periods of time. Spaceborne Computer is laying the groundwork to more quickly and successfully deploy the latest technologies for long-term space exploration and research.
At this year’s SuperComputing event in Denver, HPE received the Hyperion Research “HPC Innovation Excellence Award” and the HPCwire Editors’ Choice Award for “Top Supercomputing Achievement” based on their groundbreaking accomplishments with NASA—and now, HPE is sharing their efforts with the HPC community.
In three sessions on Birds of a Feather (BoF), experts from HPE, NASA, and other organizations discussed the status of Spaceborne Computer as well as exciting next steps. On Wednesday, Mark Fernandez, John Kichury, and I (HPE) presented “HPC in Space: Supercomputing at 17,500 MPH.” This session took a deep dive into Spaceborne Computer, exploring its unique software architecture and how it extends HPE’s presence in edge computing—literally out of this world. Following the presentation, HPE addressed attendees with two questions: “Are there additional benchmarks we should be running?” and “What would you like to run on Spaceborne Computer and why?” These names and ideas were collected, and NASA and HPE have agreed to begin onboard processing after 100–120 days, in an effort to grow the community of space exploration and ultimately push the envelope on HPC in space.
Also on Wednesday, Franke Baetke (HPE) and Sarp Oral (Oak Ridge National Laboratory) presented “Lustre Community BoF: Lustre Deployments for the Next 5 Years.” The discussion covered the mission-critical applications of Lustre and the topics at this year’s Community BoF, as well as summaries of LUG2017 and LAD 2017.
To conclude these sessions, Sven Breuner (ThinkParQ GmbH) and Franke Baetke (HPE) presented “BeeGFS – Architecture, Implementation Examples, and Future Development” on BeeGFS, an open source parallel file system that is used in several TOP500 systems.
NASA and HPE are working to expand human knowledge, from Earth to the edge of the universe, running HPC applications on the ISS. For more information on Spaceborne Computer, I invite you to follow me on Twitter at @EngLimGoh@NASA@Space_Station@HPE_HPC