HPC community prepares for new CFI competition and seeks new funding sources to complete long-range plan

Canada's high performance computing (HPC) community has identified the first of several anticipated financial commitments as part of a long-term plan (LRP) to achieve increased and stable funding for a national system of HPC research facilities. The decision last year by the Canada Foundation for Innovation (CFI) to establish a new National Platforms Fund (NPF) has spurred the development of a single unified HPC proposal for up to $60 million, which translates into $150 million once the CFI's 40-60 funding formula is applied.

The Canadian High Performance Computing Collaboratory (C3.ca Association Inc) - the consortium preparing a CFI proposal for a Spring 06 deadline - recently released its long-awaited LRP for HPC (R$, December 21/04).

Entitled, Moving Canada to the Forefront of 21st Century R&D, the LRP calls for a commitment of $76 million in public funding annually beginning in 2006 and ramping up to $87 million in 2009 and $96 million by 2012. Corresponding industrial funding for the three periods is $20 million, $22.5 million and $25 million. That compares to an estimated $250 million in funding over the past five years from government, universities and industry.

Under the LRP, Canada's seven regional HPC centres (see chart) would specialize in specific areas, but would be accessible to researchers nationwide by utilizing the CA*net 4 backbone operated by CANARIE Inc. The bulk of the requested funding would go to support mid-range HPC facilities such as the C3.ca consortia, followed by a pan-Canadian high-end facility that would be included in the world's top 30 facilities.

"The Long Range Plan solved the CFI problem of supporting HPC across Canada. It was instrumental in the creation of the National Platforms Fund," says Jacques Lyrette, chair of C3.ca's board of directors and CEO of Ottawa-based ADGA Group Consultants Inc. " Now we need an agreement to apply to the CFI as one proposal."

HPC is widely recognized as a increasingly critical component of a whole range of research activities from health, biotechnology and drug design to environmental modelling aerospace and nanotechnology. Many countries have ramped up spending to ensure that their industrial and institutional researchers are not at a competitive disadvantage. Canada currently stands at 14th globally in terms of HPC capacity (gigaflops per GDP) and implementation of the LRP would vault the nation into seventh place overall.

But the CFI funding, if secured, will only address Canada's need for mid-range regional facilities. C3.ca officials are currently exploring other funding avenues to address the financing requirements of operations, technical support staff and IMPACT Canada, the national organizing body. The CFI is also offering $18 million in operational support.

"The whole HPC community is very pleased with what the CFI has done. It addresses the hardware needs for the next three years and part of operating costs," says Dr Kerry Rowe, VP research at Queen's Univ and head of the team responsible for the LRP. "But we need a more sustainable solution. NSERC (Science and Engineering Research Canada) could fund some of the operational costs under its major facility access program but it is under enormous pressure."

Lyrette says the National Research Council (NRC) should be playing a role in supporting human infrastructure and establishing a high-end facility but it is not demonstrating much support for the venture or C3.ca. Therefore C3.ca is setting its sights on other government departments, in particular Environment Canada which is the biggest government user of HPC.

Environment Canada has a large HPC facility at Dorval PQ which is used for weather modelling and prediction, but it does not have an R&D facility.

"Someone needs to fund the people side but currently there's no candidate for this. It's the problem of finding a ball carrier within government," says Lyrette. "This is crucial to the future and it builds on the CANARIE infrastructure all across the country. The Liberals never picked it up but it's a natural for a new government to support national initiatives like this."

CFI HOLDS WORKSHOP

Late last year, the CFI held a workshop with the HPC community to outline details of the NPF's requirement for a coordinated planning approach. It was decided at the workshop to focus on mid-range computational facilities, leaving the high-end facility for another funding source. But it was also made clear that the proposal should specify the type and area of specialization for each HPC centre within the consortium.

"It's up to the community, knowing how much money is available, to come up with the architecture and strategy for their proposal," says Carmen Charette, CFI's senior VP. "The competition is between Canada and the rest of the world so how does Canada position itself to ensure competitiveness? Holistic planning is required. The CFI has never done this before so it's a bit of an experiment."

The key to ensuring a holistic approach is IMPACT Canada, the proposed management and coordinating body. It will also undertake outreach and training activities and act as an entry point for smaller firms seeking access to HPC facilities, personnel and research outcomes.

R$