Spaceborne Computer-2 sets new benchmark for AI and ML on ISS

Spaceborne Computer-2 sets new benchmark for AI and ML on ISS
by Clarence Oxford
Los Angeles CA (SPX) Feb 06, 2024

Hewlett Packard Enterprise (NYSE: HPE) has taken a significant step forward in space-based computing by dispatching the third iteration of its Spaceborne Computer to the International Space Station (ISS). This event marks a pivotal moment in the evolution of high-performance computing (HPC) capabilities beyond Earth's atmosphere, leveraging commercial off-the-shelf (COTS) servers. The latest journey of the HPE Spaceborne Computer-2 aboard a SpaceX Falcon 9 rocket, part of a Northrop Grumman Commercial Resupply Services mission contracted by NASA, signifies a leap in the scope and complexity of computational tasks that can be performed in space, encompassing artificial intelligence (AI) and machine learning (ML) workloads.

This iteration of the Spaceborne Computer, built upon the award-winning HPE Edgeline and ProLiant servers, features an unprecedented upgrade with over 130 TB of flash-based storage from KIOXIA. This enhancement not only represents the most significant storage capacity ever sent to the ISS on a single mission but also facilitates the execution of new application types and research using larger datasets through the ISS National Laboratory. The inclusion of KIOXIA 960 GB RM Series Value SAS, 1,024 GB XG Series NVMe, and 30.72 TB PM6 Enterprise SAS SSDs exemplifies the system's readiness for advanced research tasks.

The Spaceborne Computer-2's advancements are not limited to storage. Updates to the operating system, NASA space flight support software, and system security measures have been implemented, ensuring that the system's performance under the harsh conditions of space is meticulously monitored following its installation aboard the ISS.

One of the most notable benefits of this onboard supercomputer is its ability to drastically reduce the need for large-scale data downloads to Earth. By processing data in low Earth orbit, researchers can achieve a 30,000 times reduction in download size by transmitting only the essential data outputs or insights. This capability is critical for advancing Earth and space-based research, enabling near-real-time data evaluation and significantly shortening the feedback loop for scientific discovery.

The research agenda for the HPE Spaceborne Computer-2 includes conducting a federated learning (FL) experiment. This experiment aims to train ML models and inference engines that were initially developed in the cloud, independently. By collaborating with cloud service providers, the experiment seeks to enhance ML training models used on Earth while keeping an AI inference engine in space up to date. This dual-purpose initiative underscores the potential of space-based computing to contribute to AI and ML advancements both in orbit and on the ground.

Furthermore, HPE has opened the door for public engagement by inviting proposals for experiments to be conducted on the Spaceborne Computer-2 through the ISS National Laboratory. This initiative offers an unprecedented opportunity for researchers and the public to directly contribute to the frontier of space research and computing.

The deployment of HPE Spaceborne Computer-2 to the ISS represents a significant milestone in the utilization of space for advanced computational research. By leveraging cutting-edge technology and fostering collaborative research initiatives, HPE continues to push the boundaries of what is possible in space-based computing, opening new avenues for scientific discovery and technological innovation.