The Reg chats to HPE’s HPC man about NASA’s supercomputers, lunar ambitions and Columbia • The Register

By  |  0 Comments
Related Products

Interview Not content material with stashing an Apollo-based supercomputer aboard the ISS, HPE will even be slinging the petaFLOPS for NASA’s Ames Analysis Middle.

The Reg had a chat with supercomputing veteran Invoice Mannel, vice chairman and common supervisor of HPC and AI at HPE, who informed us the corporate is trying past Earth orbit.

Based mostly on Hewlett Packard Enterprise’s SGI 8600 system, the three.69-petaFLOPS beast represents the start of a four-year collaboration between NASA Ames and HPE. Named Aitken, after US astronomer Robert Grant Aitken, the machine will probably be supporting modelling and simulations of entry, descent and touchdown for the company.

It is going to additionally, after all, play a job in NASA’s upcoming Artemis programme, geared toward touchdown people on the lunar South Pole area by 2024.

Mannel was previously with supercomputing stalwart Silicon Graphics (SGI), and rose to the place of common supervisor of Compute and Storage throughout his lengthy profession there earlier than becoming a member of HPE in 2014 to move up the Excessive Efficiency Computing and AI group.

HPE, after all, snapped up SGI in 2016 for $275m.

Columbia calculations

HPE and SGI have loved a prolonged relationship with NASA, in keeping with Mannel, who informed us about his time working at NASA on SGI’s IRIS 3000 earlier than he signed up with the supercomputer firm.

He went on to recall the aftermath of the Area Shuttle Columbia catastrophe, when NASA urgently wanted a hefty supercomputer in response to what Mannel delicately described as “the Columbia state of affairs”.

The area company, Mannel mentioned, wanted the equipment to make sure the Area Shuttle programme was “wholesome”. Actually, an unlimited quantity of modelling and simulation was required earlier than the orbiters might fly once more.

We would advocate a learn of former NASA Flight Director Wayne Hale’s weblog to get a really feel for the period.

Talking with pleasure, properly over a decade on, Mannel recalled that it took simply 120 days to get the factor arrange, the place it will briefly take the number-two place on the supercomputing record. “It was nearly the primary”, he mentioned ruefully, “however on the final minute Lawrence Livermore added some compute…”

As supervisor for Altix product line, Mannel informed us the deployment was one of the vital difficult undertaken by the corporate and added that it had a “main influence” on his personal life that 12 months.

Named after Columbia, the supercomputer was ultimately decommissioned in 2013.

Going modular with NASA

Remarking that the US area company has “a really bold agenda for supercomputer functionality” – Mars is, in any case, on and off the agenda relying on who’s operating the US authorities at any given time – Mannel emphasised the flexibleness of the present method at Ames.

Describing the brand new facility as “actually only a concrete pad”, he went on to clarify the nifty design side inside it that may permit the company to “to drop containers down of various sizes and capabilities as know-how modifications over the approaching years”.

The container in query this time round comprises the HPE SGI 8600, aka “Aitken”. In addition to three.69 petaFLOPS of theoretical peak efficiency, the supercomputer additionally contains 221TB of reminiscence and 46,080 cores. Fairly a soar from Columbia.

And it’s a tad extra vitality environment friendly. “It really makes use of ambient air,” mentioned Mannel, “to chill the water that’s used to chill the supercomputer.”

“Principally, you dribble water over a pad and blow air by means of it.” Readers could be extra aware of the time period “Swamp Cooler”.

The method has enabled the supercomputer to attain a formidable Energy Utilization Effectiveness (PUE) of 1.03.

Supercomputing within the cloud?

The modular method will see NASA capable of replace its functionality through the years. Nevertheless, Mannel was lower than satisfied that the general public cloud could be pressed into service any time quickly for comparable eventualities.

Explaining that with a 99 per cent utilisation, the general public cloud would “simply be extremely costly”, Mannel cited the instance of a buyer who had spent $20m on one among HPE’s most interesting, having calculated that doing the identical within the public cloud would price nearer $200m.

“NASA did really have a look at a public cloud, versus doing the Aitken supercomputer,” mentioned Mannel, however “they discovered it was not price aggressive in any respect.”

And it would not be like NASA to waste tens of millions and even billions of taxpayer now, would it not?

To the Moon and Mars

Mannel was additionally eager to see supercomputers play a component in the way forward for human area exploration. The corporate is flush with success following the almost two-year sojourn of its Apollo (no, not that one) machine aboard the ISS.

Mannel hoped that the expertise had demonstrated to the US area company that it will possibly equip its spacecraft with extra up-to-date equipment somewhat than hardening one thing in-house that could be properly old-fashioned earlier than getting anyplace close to the launchpad.

Certainly, Mannel informed us that the gang had been getting a number of curiosity from the likes of Northrop Grumman following the success of the Spaceborne laptop. Northrop is at present anticipated to construct the habitation module for NASA’s Lunar Gateway.

Assuming the company would not change course once more.

A cynic would possibly surprise if all this computing grunt is definitely wanted. In spite of everything, the compute energy behind the missions of the 1960s and early 1970s is commonly a punchline utilized by those that maybe do not grasp the challenges concerned.

For Mannel, it is all concerning the quantity of research that might be executed on the spacecraft or habitat somewhat than having to transmit information again to Earth for processing. Having gathered telemetry from plane again within the day, he was impressed at how a lot was executed by as we speak’s computer systems on the aeroplane itself.

The identical method on a crewed spacecraft, he reckoned, “could make the mission safer and the crew safer”.

And about that 2024 aim? “I am unable to reply that… however we will probably be able to promote them laptop to assist them get there.”

It’s HPE, in any case. ®

Disrupting with information – challenges for digital native organizations


You must be logged in to post a comment Login