Updated 3:30 pm ET, Monday, April 2
By its own admission IBM isn’t exactly a name brand when it comes to astronomy. But the company aims to change that by developing software and other computing technology to run what’s projected to be the world’s largest space telescope, a massive multinational project called the Square Kilometer Array, which will generate double the data transferred across the current Internet every day.
The telescope’s name, SKA for short, describes the area (1 kilometer squared or about 0.38 square miles) of some 3,000 separate dish antennas that will make up the telescope, once construction is completed, estimated 2024.
The antennas, each about 15 meters wide (49 feet) will be arranged in a spiral extending some 3,000 kilometers from center to tip. No location has yet been chosen for the telescope, but one is expected to be announced in the coming days, and it will likely be South Africa.
By contrast, the largest current connected radio telescope, LOFAR, located in Europe, has an area of about 0.3 kilometers squared or about 0.1 square miles.
Together, SKA’s antennas will be able to peer deeper into space than any previous instrument of its kind, 100 times more sensitive than current radio telescopes, detecting the faintest, earliest signals of radiation emitted during the Big Bang, the explosion of matter that scientists believe birthed the universe as we know it.
“The objective is to figure out the really basic stuff: Why we’re all here, how we came into being, along with all of the stars, galaxies, planets and other stuff that makes up our universe,” said Dr. Ronald Luijten (pictured below), a computer scientist at IBM Research in Zurich, Switzerland, in a phone interview with TPM.
Lujiten is participating in IBM’s side of the SKA telescope project, called DOME. On Monday, IBM and ASTRON, the Netherlands Foundation for Research in Astronomy, announced DOME: a $43.8 million (32.9 million Euro) deal for IBM to develop the computer system that will power the SKA telescope. The cost of the telescope itself is estimated to be about $2 billion, or 1.5 billion euros.
The SKA telescope’s field of vision will be so great though, that the amount of data generated by its sky surveys will be almost unworkably enormous: generating 1 exabyte daily. That’s equal to about twice the amount of traffic on the Internet per day, currently.
Add to that the fact that scientists don’t want to be able to just record all this data, of course. They need to sort through it to find the most useful pieces of information, to discern the signals out of the noise, quite literally in this case, as the telescope will continuously survey all space frequencies from 7 MHz (amateur radio) through 10 GHz (weak microwaves).
“In some sense you’re looking at a needle in a haystack,” Luijten told TPM, “The signals are so small.”
The computing demands of the SKA telescope are far too great for any single system currently in operation to handle. As the SKA telescope project’s own interim director recently told CNN: “We’re talking about huge computing hardware that still needs to be developed. There’s a big challenge there.”
Enter IBM: The company, which is already one of the globe’s longtime leaders in supercomputing systems (in terms of number of systems installed and sheer processing power), now plans to take on what will be among the most daunting computing challenges in its 101-year-history.
“It is a once-in-a-lifetime opportunity for a scientist,” Luijten said. “I’ll have been in research for 27 years in a bit, and there’s been nothing like this.”
That’s not to say that IBM doesn’t see some practical commercial value in its participation in SKA. As Luijten explained to TPM, the company plans to take the supercomputing advances made in DOME and use them in its own increasingly ambitious healthcare data analysis.
“The actual content of the data [between astronomy and healthcare] is absolutely different,” Luijten said. “But it’s still big data. The storage access patterns, the question of how to make it available to scientists across the world, it’s very synergistic with IBM problems we need solved.”
Editor’s note: This article originally stated that the data generated by the SKA telescope was 10 gigabytes per second, when in fact that value refers to only final processed data.
IBM’s claim that SKA telescope will produce twice the average daily Internet traffic in raw data is based on IP traffic data previously posted at Wikipedia, citing Cisco, indicating 19,707 petabytes per month was transferred across the Internet globally in 2011. That equates to an average of 657 petabytes per day. Twice that is 1,314 petabytes, or 1. 3 exabytes. 1,000 petabytes is equivalent to 1 exabyte, the value of data that is expected to be generated by SKA.
However, TPM’s search of Cisco’s historic data and latest forecast doesn’t reveal the 19,707 petabytes/month figure.
IBM stands by the 19,707 figure, with a spokesperson telling TPM that it was provided by the SKA. An IBM spokesperson also told TPM that the antenna will actually collect between “10 and 100 times more” data, but that most of it will be filtered, correlated and processed into the value of 1 exabyte a day.
CIsco isn’t sure where the 19,707 pb/month figure came from, but a spokesperson told TPM “Over the years, there are traffic figures from past updates that remain on various websites.” Cisco’s figures of “global fixed internet” traffic from 2011 list 20,634 pb/month, “fairly close to IBM’s citation.” The Wikipedia page has been updated to reflect this value.
We’ve reached out to SKA for further information on the discrepancy and will update when we receive a response
