IBM to Help Research and Develop ‘Exascale’ Supercomputing Telescope

  • Share
  • Read Later

What has IBM’s thumbs up, millions of antennas, runs 10,000 times faster than today’s radio devices, will produce upwards of 100 times as much data per year as the Large Hadron Collier and study the beginning of time itself? If you didn’t say “Why, the Square Kilometre Array, of course!” you’re forgiven, but that’s what it’s called, surely enough — a supercomputing radio telescope to rule all radio telescopes. The only catch: It’s not due until 2024.

For the next five years, at least, it’ll be prefaced by a joint 32.9 million Euros (USD$43.9m) venture by IBM and the Netherlands-based radio astronomy institute ASTRON dubbed “DOME” to research super-fast but at the same time low-power supercomputers to be used in the international SKA project. No, not the world’s most expensive calypso-angled rhythm and blues band, but the largest, most sensitive radio telescope in existence. What does something like that need to crunch all the data it receives? Computing power to the tune of several millions of today’s most powerful computers.

(MORE: World’s Most Powerful Supercomputer Is Faster than 28,000 PS3s)

The term bandied about in today’s press release to describe that amount of data is “exascale.” That’s what you get if you step up — way up — from petascale computing: a thousandfold increase, in fact. We have petascale computers today (well, not we as in you and me, but in extremely large rooms at special facilities), and you can count the total number of active petascale systems running today on both hands. It’s projected that the first exascale computers, by contrast, won’t see light of…well, artificial research facility light, until 2018.

The goal with SKA is to produce a survey telescope powerful enough to “explore evolving galaxies, dark matter and even the very origins of the universe–dating back more than 13 billion years.” But the data generated by its collection surface area — roughly the width of the continental United States — will require unprecedented amounts of computer storage: between 300 and 1,500 petabytes annually. Contrast with the 15 petabytes per year the European Organization for Nuclear Research estimates CERN’s Large Hadron Collider generates.

“If you take the current global daily Internet traffic and multiply it by two, you are in the range of the data set that the Square Kilometre Array radio telescope will be collecting every day,” said IBM researcher Ton Engbersen in a statement. “This is big data analytics to the extreme. With DOME we will embark on one of the most data-intensive science projects ever planned, which will eventually have much broader applications beyond radio astronomy research.”

Among the solutions ASTRON and IBM are looking at to accomplish their goals: advanced accelerators, 3D stacked chips (for energy efficiency), optical interconnect technologies, nanophotonics (nano-scale light transmission), next-generation tape storage systems and phase-change memory technologies.

And there’s a “green” angle to the project as well, or at least that’s how the PR team’s pitching it: “Large research infrastructures like the SKA require extremely powerful computer systems to process all the data,” said Marco de Vos, Managing Director of ASTRON. “The only acceptable way to build and operate these systems is to dramatically reduce their power consumption. DOME gives us unique opportunities to try out new approaches in green supercomputing. This will be beneficial for society at large as well.”

Where will SKA’s millions of antennas set up shop? We’re still waiting for a final decision, due yet this year, but selection is down to either Australia or South Africa.

MORE: Meet Intel’s Crazy 50-Core ‘Knights Corner’, World’s First 1 TFLOPS Processor