NASA pulls together pieces for its most powerful supercomputer yet
Space agency plays the long game by adding performance over time... now with mix of Intel, AMD
After years of continuously upgrading a supercomputer from 2008, NASA has announced its most powerful system yet to tackle everything from upcoming Artemis missions to simulating launches, and beyond.
The thing is, you've probably already heard of it. The "Aitken" system has been inside the space agency since 2019. But with a fresh add-on, it can tackle NASA work at a scale previously impossible.
NASA builds its supercomputers a bit differently than its counterparts at national labs and big universities. Instead of commissioning a monolithic machine with a defined operational period, it takes the modular route, gradually adding more and more capacity, as we saw over the course of the last fourteen years with its "Pleiades" supercomputer, which has been operational since 2008.
This is the case with "Aitken", which is officially the space agency's most powerful supercomputer following the addition of a mere four HPE "Apollo" racks based on AMD's "Rome" architecture. The addition of AMD nodes is notable here as the system to date has been based on Intel's "Cascade Lake".
Densities being what they are, four racks is one heck of a lot of compute. The system now sits at 3,200 nodes (over 308,000 "Rome" cores) which even before the upgrade, placed the system at #58 on the Top 500 list of the world's most powerful systems. Prior to the expansion, the "Aitken" supercomputer was based on twelve (now sixteen) HPE Apollo 9000 racks and also features HPE E-Cells providing 1,152 Intel "Cascade Lake" nodes, with 40 cores per node.
"This sizeable enhancement — a 16% increase in performance since its previous expansion, and a 49% increase since last year, when the system was ranked at number 72 on the June 2021 Top500 list — translates to solving larger problems with faster results for important NASA research projects in aeronautics, space exploration, Earth science, and astrophysics," according to Michelle Moyer at NASA Ames Research Center.
NASA's "Pleiades" has a history far longer than most scientific machines and a workload track record that includes analysis of the Kepler Mission data, dark matter research, and a full-scale visualization of Earth's ocean currents, in addition to a wide range of CFD and other Jet Propulsion Laboratory (JPL) and NASA research.
In 2008, when first operational, "Pleiades" was the third most powerful supercomputer on the planet and was still ranking in the Top 50 of the Top 500 list in 2019. The long-standing system went through multiple expansions but started as an SGI-built super based on Intel "Harpertown" processors before moving through the full timeline of Intel chip releases until ending with a "Haswell" and "Broadwell" combo that brought its ultimate peak performance to just over seven petaflops.
"Pleiades" is still operational and will sit alongside its younger, faster counterpart at NASA Ames. It is housed in the first module of NASA's Modular Supercomputing Facility (MSF). The MSF site is planned to support up to 16 modules for computing and data systems.
The modules are outdoor datacenter "pods" which have both air and closed loop cooling that provides more efficient cooling than an indoor datacenter, according to NASA.
Also at the MSF is another modular system called "Electra" which is another HPE system based on the company's ICE X and E-Cell architecture. This was the forerunner to the modular systems approach at NASA.
The fact that there are 16 more modules to fill on the modular datacenter pad at NASA Ames shows we can expect NASA to keep expanding, much like the universe. ®