HPE Builds Supercomputer for NASA's Artemis Moon Mission

Hewlett Packard Enterprise (HPE) has announced Atiken, a new 3.69 PFLOPS supercomputer for NASA that will run simulations of entry, descent, and landing on the moon for NASA’s Artemis program. The supercomputer is located in a new computing and storage facility of NASA Ames Research Center.

The system is based on HPE’s end-to-end SGI 8600 high-performance computer (HPC) platform, HPE disclosed on Thursday. This is a high-density cluster with compute nodes containing second-generation Intel Xeon Scalable processors and Mellanox’s InfiniBand networking. The supercomputer has a total of 1,150 nodes and 221TB of memory. Each of these nodes has what appears to be two 20-core CPUs and no accelerators. The resulting supercomputer has a theoretical performance of 3.69 PFLOPS,k which would rank 55^th in the June 2019 TOP500 list of supercomputers.

Atiken will be located in a new Modular Supercomputer Facility in Mountain View (California) of NASA’s Ames Research Center. This facility also received its grand opening on Thursday. NASA said that the supercomputer is housed in the first module of the facility which is based on a Modular Data Center (MDC) approach that NASA jointly developed with HPE, for greater efficiency. 

The facility has the potential to expand to 16 modules for computing and data storage. Its cooling systems consist of a combination of “outdoor air, fan technology and a circulating water system” and replaces the need for a cooling tower (and millions of gallons of water), HPE said.

The supercomputer will be used for modeling and simulations of entry, descent and landing (EDL) in support of NASA’s Artemis program. This program has the goal to land astronauts (including a woman) on the Moon surface by 2024, more specifically on its South Pole.