Canada’s advanced scientific research communities are putting their hands up for more teraflops.
Teraflops, or trillions of calculations per second, are the benchmark for measuring supercomputing power. And Canada just isn’t generating enough teraflops to be internationally competitive, according to C3.ca Association Inc.
Canada’s high-performance computing (HPC) community, administered by Ottawa-based C3.ca, has called on the federal government for tens of millions in increased funding, to bring this country’s research and development up to international levels.
The organization recently released its long-range plan, outlining a national strategy for government funding in a new national high-end (tera-scale) supercomputing facility.
C3.ca is a not-for-profit organization whose members include research scientists at Canadian universities and resource providers such as Canarie Inc., which operates the advanced CA*net 4 Internet backbone.
Teraflops are all the rage in scientific research and development, but this powerful supercomputing tonic comes with a potent price tag. The strategy seeks federal funding of $76 million per year from 2006 (plus $20 million from industry), growing to $97 million (plus $25 million) per year in 2012 and beyond.
IBM Corp.’s eServer Blue Gene Solution currently holds the top two positions in world teraflop supercomputing powers, with U.S. organizations taking the top six rankings. Japan, Spain and the Netherlands also crack top-10 positions, but Canada isn’t even in the top 50.
“We basically don’t have a high-end research facility in Canada,” said Jacques Lyrette, chair of C3’s board of directors. “We don’t have those facilities for research that exist elsewhere in the world, in Europe and in the U.S.”
But it’s not so much the building of the tera-scale facility, as its sustained operation and technical support that will crunch the dollars, says Lyrette.
The long-range plan encompasses people funding as a core aspect, to support technical expertise at the new high-end facility, as well as at Canada’s existing six mid-range HPC facilities: WestGrid in western Canada, SHARCNET and HPCVL in Ontario, CLUMEQ and RQCHP in Quebec, and ACEnet in the Atlantic provinces.
Funding until now has come from mainly the Canada Foundation for Innovation (40 per cent) and the universities and industry members (60 per cent), says Bruce Attfield, secretary and treasurer for C3.ca, but there’s been no government funding to support the growth of highly qualified HPC personnel.
“A researcher has to stop and learn how to use these sophisticated machines, how to get the most benefit out of them,” he said. “Funds haven’t been forthcoming to ensure sufficient people to manage the software and to cover the high costs of operating these facilities.”
Researchers are doing the day-to-day running of these computers, says Lyrette. “The human resources to run these beasts are not being financed. We need people who specialize in maintaining HPC, so our researchers can go back to researching.”
C3.ca says more than $250 million has been invested or committed to high-performance computing in Canada over the past five years, by the federal government, the provinces, universities and its industry members. C3.ca also receives funding and technical support from seven HPC industry vendors – Apple Canada, Cray, Hewlett-Packard (Canada), IBM, NEC Solutions, Silicon Graphics and Sun Microsystems of Canada.
“There has been, over the past 10 years, increasing investment and much stronger support for a national approach to this very expensive computing,” said Attfield. “Looking at the next stage, there never were guarantees that the money would be available. So the long-range plan really is looking for sustainability.”
Universities, government and industry use these supercomputing systems for research in areas such as drug development, astrophysics, financial modeling, disease containment and nanotechnology. Having the best possible computing facilities and highly qualified personnel will significantly enhance research productivity, says C3.ca, as well as reduce time to manufacture and market, facilitate knowledge discovery and accelerate innovation.