The US National Aeronautics and Space Administration (NASA) and Hewlett Packard Enterprise (HPE) said on 22 August they have developed a custom-designed supercomputer to run simulations for the agency’s Artemis program, a mission to land the next humans on the lunar South Pole by 2024.
Named after Robert Grant Aitken, an American astronomer specializing in binary star systems, the supercomputer will help NASA’s Ames Research Centre to model and simulate entry, descent, and landing (EDL) for Artemis and other missions.
“Aitken” – which will run thousands of complex simulations more quickly at 3.69 petaFLOPs of theoretical performance to enable accurate and safe landings on the moon – is an initial development of a four-year, multi-phase collaboration between HPE and NASA Ames.
The supercomputer is based on an end-to-end, purpose-built high-performance computing (HPC) platform, which includes special liquid cooling capabilities for optimal energy efficiency, HPE said in a statement.
Aitken is located in NASA Ames’ new modular supercomputing facility, based on a Modular Data Centre (MDC) approach jointly developed with HPE, to deliver advanced HPC solutions that drive greater efficiency and significantly reduce electricity and water use.
The new facility, based in Mountain View, California, will combine native Bay Area temperature and evaporative methods to cool the supercomputer, replacing the need for a cooling tower and millions of gallons of water.
“HPE has a longstanding collaboration with NASA Ames, and together, we continue to build innovative HPC technologies to fuel space and science discovery that increase overall efficiency and reduce costs,” Bill Mannel, vice president and general manager, HPC and AI, at HPE, said.
“We are honoured to have designed the new Aitken supercomputer and power capabilities for humanity’s next mission to the moon,” he said.
Design Specs
HPE designed the NASA Ames’ new supercomputer using the end-to-end, purpose-built HPE SGI 8600 system that integrates compute, software, networking and other IT infrastructure solutions from its robust ecosystem of partners, including:
- 2nd Generation Intel® Xeon® Scalable processors for advanced compute performance
- Mellanox InfiniBand to enable scalable bandwidth for high-performance networking
- Schneider Electric SmartShelter Containers that enable easy-to-deploy, prefabricated IT infrastructure packaged within a secure, weather proof, fire-rated, data module for remote or special applications
Other features include:
- 1,150 nodes, 46,080 cores, and 221 TB of memory.
- 3.69 petaflops of theoretical peak performance.
- Power Usage Effectiveness (PUE) of 1.03.
Leave a Reply