U of T does the math on Big Blue supercomputer

When you’re dealing with complex research areas such as theoretical astrophysics or computational biology, using any old research system isn’t going to cut it, according to the University of Toronto. That’s why the school has teamed up with IBM Corp. to create an energy-efficient supercomputer that aims to process over 360 trillion calculations per second.

The university’s SciNet Consortium – which undertakes research in aerospace, astrophysics, climate change prediction and medical imaging – said the supercomputer will combine IBM’s Power6 architecture with Intel Corp.’s soon-to-be-released Nehalem processor. The hybrid system will feature more than 4,000 servers linked together and will be connected to a five petabyte storage complex.

Funding for the project was provided by the Canadian Foundation for Innovation’s National Platforms Fund, in partnership with the Province of Ontario and the University of Toronto.

“SciNet is basically the largest collection of research scientists in the country and we’ve always had the need for a very powerful computing system,” Richard Peltier, scientific director of SciNet and the head of the Centre for Global Change Science at the university, said. “This will bring easily the most powerful computing system that the country has ever owned into Canada.”

Peltier said the new system’s power is equivalent to 30,000 desktop computers and will be roughly 100 times faster than the machines currently in use at the university. This will especially come in handy, he said, during his climate change and planetary atmosphere research.

“We’ll be able to integrate models which are much more highly resolved spatially than those which we’ve been able to use in the past, so it will have a huge impact on our ability to do climate system simulation,” he said.

Neil Bunn, IBM’s lead architect on the system, said the system will not only benefit University of Toronto researchers, but also boost Canada’s profile and contribution to global science. The impact the new computer will have on Peltier’s climate change research alone will provide huge dividends, he said.

“He’ll now be able to run many simulations that before would have taken months to do,” Bunn said. “He can determine what happens if we reduce our greenhouse gases by various amounts, what happens if we change fuel sources, or what happens if different changes occur to the various ice flows in the world.”

But along with the supercomputer’s incredible power and storage capability – which IBM estimates will be able to store 60 times more data than the Library of Congress Web archive – the university is also pushing for an energy efficient system. Bunn said that the system will require a substantial amount of energy to operate and estimated it would consume upwards of two megawatts of power or 2,000 kilowatts an hour. Where the system will differ from other power hogging supercomputers, he said, is that IBM plans to use an innovative watercooling system to keep cooling costs to a minimum.

“Most systems are built in data centres and you spend as much power cooling the system as you do actually running it,” Bunn said. “Because we’re going to use water, for every one kilowatt of power that we use, we’ll only be using 200 watts or less for cooling power. That’s an efficiency ratio that’s never been done before.”

According to the university, the green aspect of the build was just as important as its technical capabilities.

“This was extremely important to us, especially for myself as an environmental researcher,” Peltier said. “We expect that this machine will be one of the most energy efficient in the world.”

In another effort to keep costs down and to avoid the need to build a costly downtown power substation, the supercomputer will be built and housed north of the university in Vaughan, Ont. A 14,000-square-foot warehouse has been rented to hold the machine, which is scheduled to be completed next summer.

Related Download
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center Sponsor: Lenovo
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center
Find out how Hyperconverged systems can help you meet the challenges of the modern IT department. Click here to find out more.
Register Now