Personal blog of Yzmir Ramirez

NASA and Hewlett Packard Enterprise (HPE) have teamed up to build a new supercomputer, which will serve NASA’s Ames Research Center in California and develop models and simulations of the landing process for Artemis Moon missions.

The new supercomputer is called “Aitken,” named after American astronomer Robert Grant Aitken, and it can run simulations at up to 3.69 petaFLOPs of theoretical performance power. Aitken is custom-designed by HPE and NASA to work with the Ames modular data center, which is a project it undertook starting in 2017 to massively reduce the amount of water and energy used in cooling its supercomputing hardware.

Aitken employs second-generation Intel Xeon processors, Mellanox InfiniBand high-speed networking, and has 221 TB of memory on board for storage. It’s the result of four years of collaboration between NASA and HPE, and it will model different methods of entry, descent and landing for Moon-destined Artemis spacecraft, running simulations to determine possible outcomes and help determine the best, safest approach.

This isn’t the only collaboration between HPE and NASA: The enterprise computer maker built for the agency a new kind of supercomputer able to withstand the rigors of space, and sent it up to the ISS in 2017 for preparatory testing ahead of potential use on longer missions, including Mars. The two partners then opened that supercomputer for use in third-party experiments last year.

HPE also announced earlier this year that it was buying supercomputer company Cray for $1.3 billion. Cray is another long-time partner of NASA’s supercomputing efforts, dating back to the space agency’s establishment of a dedicated computational modeling division and the establishing of its Central Computing Facility at Ames Research Center.

Last year, HPE successfully built and installed a supercomputer on the International Space Station that could withstand the rigors of being in space. Today, the company announced that it is making that computer available for earth-based developers and scientists to conduct experiments.

Mark Fernandez, who has the lofty title of America’s HPC Technology Officer at HPE, says that the project was born with the idea that if we eventually go to Mars, we will need computers that can withstand the travel conditions of being in space for extended periods of time.

What’s more, because space computers have traditionally lacked the sophistication of earth-based computers, they conduct some of the work in space and then complete the calculations on earth. With an eye toward a Mars trip, this approach would not be feasible due to the distances and latency that would be involved. They needed a computer that could handle processing at the edge (in place) without sending data back to earth.

The original idea was to build a supercomputer with the state of the art off-the-shelf parts as and install it on the ISS as an experiment to see if this could work. They built the one teraflop computer in the summer of 2017 and launched it into space on a SpaceX rocket. The computer was built with Intel Broadwell processors, which Fernandez says were the best available at the time.

The first step was to see if the computer they built could handle the launch, the cold temperatures of waiting to be on-boarded, the solar radiation and generally uncommon conditions of being in space.

Once installed, they needed to figure out if this computer could operate in the power and cooling environment available onboard the ISS, which is not close to what you would have in earth-based datacenter with a highly controlled environment. Finally, once installed, would the computer operate correctly and give accurate answers.

The special sauce here was a package of software they call Hardened with Software. “We wrote a thin, lightweight way suite of software to quote-unquote, harden our systems of software, so you can take state of the art with you,” he said.

The computer was launched in August 2017 and has been operating ever since, and Fernandez says that it has worked according to plan. “So we’ve achieved our signed, dated and contracted mission. We have a one teraflop supercomputer on board the International Space Station with Intel Broadwell processors.” He says that supercomputer has flown around the earth 6000 times since launch.

The company now wants to open this computer up as a kind of service to earth-based developers and scientists to experiment with high-latency jobs that would have required some processing on earth. With the HPE Spaceborne Computer available to use, they can see what processing this information at the edge would be like (and if it would work). The computer will be in operation until some time next year, and in the meantime interested parties need to apply to HPE and NASA to get involved.

 President Obama recently re-affirmed America’s commitment to sending a manned mission to Mars. Think your data science challenges are complicated? Imagine the difficulties involved in mining data to understand the health impacts of a trip to Mars. When sending humans “where no one has gone before,” there are a multitude of variables to consider, and NASA is hard at work… Read More