Quebec universities prep for high-performance computing

The leader of a Quebec high-performance computing consortium, awarded $8 million on Tuesday, said funding from the Canadian Foundation for Innovation (CFI) and provincial governments is establishing a computing standard that will be a challenge to maintain once initial monies are spent.


Wagdi Habashi, a McGill University mechanical engineering professor and project leader of CLUMEQ, a consortium joining the University of Laval, the University of Quebec in Montreal and McGill as well a trio of Eastern Quebec institutes, said high performance projects across Canada will in three or four years need further investment from businesses or the CFI.

“”By the time you install it, there are better machines. And if you want to continue at the supercomputing level, you have to continue your investment,”” Habashi said. “”In the long run, what is missing is maintaining these things. I don’t think you can sit at the university and generate the type of funds that will allow a system like that to be updated.””

The announcement included for CLUMEQ a $4 million grant from the CFI and matching funds from the Quebec government for a total of $8 million that Habashi will use to establish a joint high-performance computing facility for CLUMEQ’s member institutions. Most of the equipment — a 64-processor shared memory computer and a 256-processor distributed memory cluster — will be centred at McGill, though each institution will house some of the high-performance technology and each will be able to access all the shared resources.

“”You have 256 CPUs that are interconnected and you can access them all at once,”” Habashi said. “”Theoretically, you could do your job 256 times faster.””

The facility will be employed for research in aerospace and other engineering, nanomaterials, bioinformatics, environmental flows and computer science.

“”When you analyze a plane for performance, you analyze components separately because you can’t analyze it all at once,”” said Habashi, also a research fellow for Longueuil, Que.-based aviation engine maker Pratt and Whitney Canada Corp. “”When you get that (high-performance) type of power, you analyze the entire plane. You see problems you wouldn’t have seen before.””

Habashi said the increased computational power allows researchers to virtually analyze in-air maneuvers and the effects of extreme weather on a plane’s peformance. Advanced computing power, he said, can eliminate the need for expensive testing in wind tunnels.

The inter-university sharing of resources technically qualifies the CLUMEQ project for grid computing status, though true grids share more than just computational cycles, according to Andrew Bjerring, president and CEO of the Canadian Network for the Advancement of Research Industry and Education (Canarie) Inc.

In their most advanced sense, grids have the ability to automatically spread and collect their users’ computing resources, from computer cycles to computer applications on an as-need basis, in much the same way an electrical grid manages electricity. Though that so-called service grid model is at least few years away, the CLUMEQ project is one of a number of Canadian university “”partner grids,”” wherein institutions share resources, including both computational cycles and applications.

The list of Canadian partner grids includes Westgrid in Alberta and British Columbia and Sharcnet and the High Performance Computing Virtual Laboratory grid in Ontario. Bill St. Arnaud, Canarie’s senior director of network projects has predicted a functioning grid stretching from coast to linking through Canarie’s Ca*net 4, a 40 Gbps optical network that puts light paths within the user’s control, within one year. Habashi said CLUMEQ plans to be part of the national grid.

“”What is happening in Canada is a huge change from four or five years ago, and it’s all been caused by the CFI,”” he said. “”Before the CFI came in, super-computing was a dirty word in Canada. It didn’t matter where you applied (for funding), you were always rejected.””

This skepticism about high-performance computing can be traced back to a pair of projects at the University of Toronto and the University of Calgary in the 1980s, according to Bjerring. “”Neither of those facilities was able to last much beyond the initial round of government funding,”” he said. “”They also had industry funding, but nowhere near enough. When the funding ran out, they had to close their shops.””

The birth of the CFI in 1997 delivered a new confidence in and new money for high-performance computing and Andrew Pollard, chair of high-performance computing advocacy group Association Inc., is hopeful the initial investments will spur more growth.

“”As a result of the CFI and provincial governments and high-performance computing vendor contributions, we now have over $160 million invested in high-performance computing in this country,”” said Pollard, also a computational fluid dynamics professor at Queen’s University. “”The community hopes investment will be long enough to create intellectual wealth that will allow more people to accomplish things that will enforce that this is an infrastructure that needs to be supported.””

Bjerring is optimistic the research will continue, but he said public money will have to figure into the equation.

“”The ability to acquire high-performance facilities has always been dependent on funds from CFI or NSERC (National Sciences and Engineering Research Council of Canada). You’ve never had a steady budget where you could set aside replacement money,”” he said. “”My crystal ball is probably no better than anyone else’s but I think there will be a need for special funding.””

Habashi suggested the survival and progression of high-performance computing might require more simply more CFI funding.

“”We have to think outside the box,”” he said. “”It could be consolidating these sites into a national supercomputing centre or centres,”” he said.


Share on LinkedIn Share with Google+