The Globe and Mail: Canada’s monster computer roars to life

June 18, 2009 in in_the_news, news

Link to Globe and Mail article on Scinet

Thursday June 18 2009:

The Globe and Mail: Canada’s monster computer roars to life

It has taken a year and $50-million to put together, and its brain takes up as much room as a warehouse full of refrigerators.

Today, the monster finally opens its eyes, as the University of Toronto’s newest supercomputer – the fastest such machine in Canada – goes online.

There’s no shortage of beastly metrics by which this computer’s power can be measured: It can perform more than 300 trillion calculations a second, simulate the Earth’s climate 100 years into the future in four days and help researchers study cosmic background radiation, a calculation-intensive task that offers a glimpse into what the universe looked like 13 billion years ago.

“This positions us on a world research stage at a whole new level,” said Chris Pratt, strategic initiatives executive at IBM Canada. “This isn’t one step or two steps; this is like, ‘Wow.’”

A small part of the IBM System x iDataPlex server has been operating since late last year, humming away in a Vaughan-area warehouse. But today, the machine’s full power is unleashed.

Almost everything about the system sounds improbable.

It uses the same amount of energy, at peak consumption, as 4,000 homes. It is about 30 times more powerful than the next-fastest research computer in Canada. It can whirl data through its digital veins at the rate equal to about two DVD movies a second. It is among the 15 fastest computers in the world, and the fastest outside the United States.

Or think of it this way: If you’ve purchased a decent home computer lately, it may have come with a relatively fast 2.53 gigahertz processor. Or maybe you shelled out more for a fast, top-of-the-line “quad core” system, which runs on four such processors.

U of T’s new toy runs on 30,240 of them.

For a system that’ll suck up at least $1-million worth of energy a year, it seems odd to describe the computer as energy efficient, but, in the superlative-laden world of supercomputers, it is.

The computer monitors its individual units continuously, and shuts off any that aren’t in use for more than about 10 minutes.

IBM’s designers also came up with what seems like a pretty obvious way to reduce cooling costs: let the Canadian winter do it. Every time the outdoor temperature drops below a certain point, the computer uses the cold from the air to chill out. In total, IBM estimates the same supercomputer’s energy use would be equivalent to that of about 750 more homes if it were designed using traditional methods.

“It’s a really impressive, world-class facility,” said Chris Loken, the chief technical officer for the SciNet consortium, which is responsible for getting the supercomputer built. “It’s going to give a lot of scientists access to some really powerful resources.”

The supercomputer’s uses, Mr. Loken notes, vary from planetary physics to aerospace research to medicine. For example, the machine will act as a data centre for the Large Hadron Collider – the world’s largest atom smasher, located underground near Geneva – which can generate more than 40-million collisions a second. Researchers are already using some of the computer’s capacity to study cosmic background radiation.

But even for a computer as powerful and expensive as this one, its reign among the world’s heavyweights will be short-lived. Both Mr. Loken and Mr. Pratt admit that, given the pace at which technology is moving, U of T’s supercomputer will be eclipsed by several faster machines in the next few years.

“It will not be in the top 20 systems next year,” Mr. Loken said. “But there’s still going to be a lot of awfully good research that can be done on it.

“There’s really good science being done now on number 100 and number 500.”

The Toronto Star: U of T supercomputer probes origins of the universe

June 18, 2009 in in_the_news, news

Link to Toronto Star article on SciNet

Jun 18, 2009:

U of T supercomputer probes origins of the universe
Go on. Ask this IBM System x iDataPlex to do as many calculations as there are stars in our Milky Way Galaxy.  Hold on. It won’t take a second. Not even close to one.Although its name may be ungainly, the University of Toronto’s new supercomputer performs so elegantly it can churn through 300 trillion pieces of information in the time it took Jamaican sprinter Usain Bolt to run 10 metres at a Toronto track meet last week.And with a measly 200 to 400 billion stars in the Milky Way, galactic-sized calculations will be child’s play.But the new computer – which vaults to the position of Canada’s most powerful upon having its last piece fired up today – will be charged with solving both astronomical and earthbound problems.

“This is to computing what the CN Tower was to architecture in Canada,” says Chris Pratt, an executive with IBM Canada.

“It has the ability to simulate and predict about 1,000 years of the Earth’s climate in about four days.”

Built for U of T’s SciNet Consortium, which includes the school’s research hospitals, the computer will be working on everything from medical imaging and the likely progress of climate change to the forces at play as the universe dawned some 13 billion years ago.

The system, which began operating in stages last year, puts some 30,240 of the world’s most powerful Intel processors together in 45 file-like stacks. It can run as many of those processors as required.

And, according to the latest TOP500 ranking of supercomputers, it enters full service as the world’s 12th most powerful.

That ranking has already helped SciNet attract world-class research, including a share of the work on the origins of the universe that will be conducted by the Large Hadron Collider project in Geneva in September. The collider will smash protons together at near-light speeds, to try to emulate conditions around the time of the big bang.

The new Toronto computer won a prestigious place among a select group of similar facilities across the globe that will try to find the big bang signatures among billions and billions of such daily collisions.

It will also be used to simulate protein creation and interactions, and help crunch the numbers on ice cap melting and weather conditions that will come with climate change.

Such computing prowess requires a correspondingly impressive input of energy – enough to power as many as 4,000 homes, Pratt says. The resulting heat generation would fry both the computer and the building in which it’s housed, but for the computer’s unique, water-based cooling system.

The system pipes water via tubes strung throughout the computer, often down to the microchip level, and dissipates it through heat exchangers on the roof.

CBC News: Toronto team completes Canada’s most powerful supercomputer

June 18, 2009 in in_the_news, news

CBC News article on SciNet

Thursday, June 18, 2009:

Toronto team completes Canada’s most powerful supercomputer
A supercomputer that can complete more than 300 trillion calculations per second — the most powerful in Canada — has been completed at the University of Toronto’s SciNet facility.

The IBM iDataPlex cluster computer would be one of the top 15 most powerful supercomputers in the world, based on the online Top 500 list, said a release Thursday announcing the supercomputer’s completion.

The project was a collaboration between SciNet, IBM Corp. and Compute Canada, an umbrella group that represents academic high-performance computing groups across the country.

Researchers hope to use the machine’s vast computing power to:

  • Analyze data from the Large Hadron Collider, the world’s most powerful particle accelerator.
  • Analyze high-resolution climate change models, including ones that predict the decrease in Arctic sea ice and regional climate change predictions from Ontario and the Great Lakes watershed region.
  • Conduct research on aerospace, astrophycis, bioinformatics, chemical physics, medical imaging, and the ATLAS research project on forces that govern the universe.

The cluster uses 30,240 Intel processor 5500 series 2.53 GHz processor cores and is cooled using water.


May 26, 2009 in in_the_news, news

Canada’s Fastest Supercomputers: Daniel Gruner and Chris Loken, SciNet Consortium

SciNet is a new High-Performance Computing consortium based at University of Toronto which operates Canada’s two fastest supercomputers in a new, highly energy-efficient datacentre. We will discuss the challenges of designing, building and operating a cluster with 30,000 cores and the surrounding infrastructure as well as the software systems.


IBM: Canada’s most powerful Supercomputer

May 8, 2009 in in_the_news, news

Learn about how IBM and the University of Toronto’s SciNet Consortium built Canada’s most powerful and energy efficient supercomputer in this video created by IBM.
Video located on the right side of page:
The video does not work in Firefox on all machines, please use Internet Explorer if you’re having trouble viewing.

SciNet and the Intel Xeon 5500

April 2, 2009 in in_the_news, news

“The Canadian Scinet machine is one of the first to use the newly launched Intel Xeon 5500.  The  Xeon 5500 is considered to be Intel’s most significant chip advance in more than a decade.”
For the full story go to:

IBM Maintains Top Spot

November 17, 2008 in in_the_news, news

From HPCwire article: IBM Maintains Top Spot on Supercomputer List

November 17, 2008

IBM Maintains Top Spot on Supercomputer List

ARMONK, N.Y., Nov. 17 — For a record-setting ninth consecutive time, an IBM system took the No.1 spot in the ranking of the world’s most powerful supercomputers. The IBM computer built for the “roadrunner project” at Los Alamos National Lab — the first in the world to operate at speeds faster than one quadrillion calculations per second (petaflop) in June 2008 — remains the world speed champion.

IBM is also leading the move to design all-new hybrid systems — such as the roadrunner project — that combine different types of processors for better performance and energy efficiency. IBM is currently building a 360-teraflop hybrid cluster for the University of Toronto, for example, a deal which will pair one of the world’s largest Power6 clusters with IBM’s new iDataPlex platform (x86) to create an extremely flexible 4000-node supercomputer capable of running a diverse range of software at high levels of performance. Starting in 2009, Canadian scientists plan to use the system to create new methods of medical imaging, among other uses. Part way through installation, the computer has already debuted at No.53 on the new TOP500.

University of Toronto to Acquire Canada’s Most Powerful Supercomputer from IBM

August 14, 2008 in in_the_news, news

System will process up to 360 trillion calculations per second, store 60 times more data than the Library of Congress Web archive, and be housed in an ultra energy-efficient data centre

TORONTO, 14 August 2008: The University of Toronto’s SciNet Consortium and IBM (NYSE: IBM) today announced an agreement to build Canada’s most powerful and energy-efficient supercomputer.
The consortium, which includes the University of Toronto and associated research hospitals, will enhance SciNet’s competitive position in globally important research projects. These include ground-breaking research in aerospace, astrophysics, bioinformatics, chemical physics, climate change prediction, medical imaging and the global ATLAS project, which is investigating the forces that govern the universe.
Capable of performing 360 trillion calculations per second, the supercomputer will pioneer an innovative hybrid design containing two systems that can work together or independently, connected to a massive five petabyte storage complex. Because it is a hybrid, the machine is extremely flexible, capable of running a wide range of software at a high level of performance.
As a premier academic research system, the machine will be among the top 20 fastest supercomputers in the world; 30 times faster than the peak performance of Canada’s current largest research system. It also represents the second largest system ever built on a university campus, and the largest supercomputer outside the United States.
“The University of Toronto has partnered with IBM to become one of the world’s premier computational research institutions – a collaboration that will attract researchers from around the world,” said Dr. Richard Peltier, Scientific Director of SciNet and Director of the Centre for Global Change Science.
As a physicist whose interests are focused on planetary physics and climate change prediction, Dr. Peltier’s work includes research on the impacts of greenhouse gas-induced global warming, which will be greatly enhanced by this system. The SciNet facility will be one of the world’s most advanced supercomputers for analyzing high-resolution global models to predict future risks, such as the accelerating decrease in Artic sea ice. An immediate project will be the construction of regional climate change predictions for the Province of Ontario and Great Lakes watershed region.
Another area of research for this system will be to explore the modern scientific mystery of why matter has mass and what constitutes the mass of the universe. Beginning in September, the Large Hadron Collider project based in Geneva, the most powerful particle accelerator ever built, will produce vast quantities of data, which scientists hope will be begin to unlock these mysteries. SciNet’s computing power and storage capacity will be a significant contributor to the data analysis.
“SciNet will have one of the best facilities in the world that will allow Canadian physicists to participate in the adventure of the Large Hadron Collider,” said Dr. Pierre Savard, a member of the Canadian group working at CERN, Geneva. “This research may change our view of matter and the universe.”
This facility will involve the largest implementation of IBM’s highly efficient iDataplex system, which holds twice as many processors per unit as standard systems and is entirely water cooled. More than 4,000 servers will be linked together in this multi-platform solution, including one of the world’s largest POWER6 clusters and Intel x86-based clusters. This IBM supercomputer will be one of the first systems to use a new processor technology from Intel called Nehalem, being introduced this fall.
“A system this complex could only be designed by bringing together the best minds from the University of Toronto and IBM,” said Chris Pratt, Strategic Initiatives Executive, IBM Canada. “This is a tremendous example of public and private collaboration that will benefit the Canadian research community for many years to come.”
Funding has been provided by the Canadian Foundation for Innovation’s National Platforms Fund, in partnership with the Province of Ontario and the University of Toronto.
Datacentre construction will begin immediately at a facility just north of Toronto. Installation of the system will begin in the fall with several milestones throughout the winter. It is anticipated that both of the main computing systems will be fully operational by summer 2009.