Long Range Plan for High Performance Computing for Canada 24 March 2003

1
Long Range Plan forHigh Performance Computing
forCanada24 March 2003
2
Background (1)

• High Performance Computing (HPC) in Canada has a
  checkered past
• ? poor funding, lack of communication, little
  coordination, no vision, inadequate personnel
  support
• C3.ca began in 1997 with the support of a few
  universities committed to improving Canadian
  researcher access to HPC resources
• Canada Foundation for Innovation came along at
  just the right time
• Regional HPC consortia have emerged to serve many
  researchers' needs
• ? MACI / WestGrid, Sharcnet, HPCVL, Clumeq, AC3,
  individual sites

3
Background (2)

• C3 membership has grown to include almost every
  University in Canada
• HPC facilities support gt1000 researchers in
  computational science, engineering, health
  sciences (including medicine), humanities and the
  social sciences
• Computational science is now an equal with
  experimental/observational and theoretical
  research paradigms and deserves similar
  recognition
• The HPC communities understand the importance of
  HPC to research and the long-term implications
  for Canada

4
Background (3)

• HPC is a cost-effective shared research facility!
  
• As a community we are frustrated by
• Exciting research opportunities, but without the
  computational tools that are needed
• Computing hardware, but without the money to
  properly operate it
• Continual cycle of proposals to CFI, NSERC,
  provinces, etc.
• Uncertain and inadequate funds to support HPC
• Lack of a coherent policy for the support of HPC
• Competing with researchers in other countries
  with better access to resources than we have
• To be competitive on the world stage, we must
  resolve these issues

5
Long Range Plan for Canadian HPC

• Create a long-term vision for HPC in Canada
• ? Understand the needs of the research community
  now and in the future
• ? Anticipate the needs and build towards
  providing the infrastructure to meet those needs
• ? Gain recognition that HPC computing and data
  resources are part of basic core infrastructure,
  in much the same way that we now take the network
  for granted
• LRP process is national effort to gain broad
  input about needs, opportunities and benefits of
  HPC to Canadas research communities
• Engage the community to come up with a united
  vision for Canada

6
The LRP Outcomes are Critical to our Sponsors
7
Authors Panel Membership

• Kerry Rowe (Chair) VP Research,
  Queens University
• Jonathan Borwein Mathematics, Simon
  Fraser University
• Russell Boyd Chemistry,
  Dalhousie University
• Gilbert Brunet Meteorological
  Services of Canada
• Hugh Couchman Astrophysics,
  McMaster University
• Alan Evans Brain Imaging, McGill
  University
• Martyn Guest Daresbury
  Laboratory, England
• Ian Lancashire English, University of
  Toronto
• Jonathan Schaeffer Computing Science,
  University Alberta

8
Advisory Committee Membership

• Andrew Pollard (Chair) Mechanical
  Engineering, Queens University
• Peter Poole Physics, St. Francis Xavier
  University
• Andre Bandrauk Chemistry, Universite de
  Sherbrooke
• Bruce Attfield President C3.ca
• Guy Felio NRC
• Kate Wilson NSERC
• David Moorman SSHRC
• Walter Stewart Silicon Graphics
• Hala Saab Intel
• TBD CIHR

9
Meeting Times and Agenda

• Local meetings and Regional Town Hall meetings
  (schedule on web)
• Agenda
• Overview of purpose, expected outcomes
• Pose questions to seek input and comment
• Identify research areas and interest in HPC
• Receive submissions (paper or web)
• Expected timeline
• MayJune 2003 draft report
• Summer 2003 for consensus building
• November 2003 for printing of final report

10
Need Your Help to Address Issues Such As

• Is access to HPC resources -- computing, data,
  networking -- for research critical to the
  Canadian community?
• Does Canada need a national strategy and
  long-range plan for HPC?
• How should long-term HPC resources be funded?
• What should a national strategy include?
• HPC is not just equipment people are an integral
  part of the success of any HPC facility. What
  level of support is needed?
• How important is the grid today (or soon) for
  your research?
• How should Canada ensure internationally
  competitive high-end facilities?

11
HPC Long Range Plan Draft Table of Contents

• Preface
• Executive Summary
• HPC A critical enabling technology Pyramid of
  computing capabilities Where Canada stands
  today HPC Vision for Canada
• Application case studies
• HPC Strategy for the future PACT -- People,
  Applications, Communications, Technology
  Budgets Timelines Milestones
• Economic and competitive benefits and impact if
  HPC needs are either met or NOT met
• Policy discussion
• Appendices

12
Communications

• Your input is valued and is critical to our
  collective success
• Please forward any comments, either in writing,
  email, or via the web to
• C3.ca Association Inc.,
• c/o CCO, Institute for Information Technology
• National Research Council of Canada
• Bldg M50
• Ottawa, ON
• Attn Ms. C. Tourigny
• www.c3.ca

13
Some questions for discussion

• Is access to HPC resources critical to the
  Canadian research community?
• how does Canada benefit from providing the
  Canadian research community with access to
  state-of-the-art HPC resources?  Can you give
  examples
• What are the short-term and long-term
  consequences if Canadian researchers do not have
  access to HPC resources?

14
Basic technology model and proposed scope of LRP

• a national facility (i.e. one that would rank in
  the top 30 in the world). Today this would be a
  Tera-scale machine.
• a number of regional facilities (each ranking,
  perhaps, between 150-500 in the world)
• local facilities (including facilities that
  could be operated by an individual, single lab,
  or even single University).
• LRP to deal with 1 2

15
Some questions for discussion (Cont)

• Does Canada need a national strategy
• and long-range plan for HPC?
• Is the proposed scope for the LRP
• appropriate?
• If so, how do we define the HPC needs for the
• Canadian research community?

16
Some questions for discussion (Cont)

• Should a national strategy include
•  a long-range funding plan for HPC? If so, over
  what period?
• What are the arguments for co-ordinated
  funding of HPC?
• at least one world-class HPC facility (i.e.,
  something that might be in the world Top 30 list
  of sites)? If so, do you have a need to use it?
  For what? How should it be managed?
• support for fostering and growing regional HPC
  facilities?

17
Some questions for discussion (Cont)

• How should long-term HPC resources be funded
•  from existing budgets (e.g. from existing
  NSERC, CIHR, SSHRC money)?
• from Canada Foundation for Innovation? Is the
  status quo acceptable or does CFI have to make a
  bigger (or lesser) commitment to HPC?
• from a new pool of funds allocated to HPC?
• by a combination of items above?

18
Some questions for discussion (Cont)

• HPC is not just equipment people are an integral
  part of the success of any HPC facility.
• Are you satisfied with the current level of
  people support for HPC (i.e. system admin.,
  consultants, programmer/analysts, data
  visualization, etc.)?
• Do we need an increased level of people support
  for HPC facilities? (If so, what?)
• Do we need to increase the level of HPC training
  of HQP in Canada? If so, why? In what areas?

19
Some questions for discussion (Cont)

• The notion of both computing and data grids is
  attracting a lot of attention.
• What do you see as the role of the grid in
  Canada?
• How important is the grid today (or soon) for
  your research?
• How important will the grid be in the long-term
  to your research?

20
Some questions for discussion (Cont)

• Technology provides opportunities for economic
  growth, thus with respect to HPC
• Which technologies depend for their growth on
  abundant, pervasive, access to HPC resources?
• Which technologies can absorb a scientific
  advance, breakthrough, result very quickly when
  supported by large HPC capability?
• Which technologies are only enabled by the
  presence of HPC resources?

21
Some questions for discussion (Cont)

• Are the regional consortia (WestGrid, SHARCNET,
  HPCVL, RQCHP, CLUMEQ, AC3, etc.)
• meeting the current needs of Canadian researchers
  with respect to HPC? 
• suitable for providing researchers with
  computing at the middle layer of the capability
  pyramid on an ongoing basis?
• providing sufficient coverage and breadth of
  resources to researchers in Canada? (If not how
  could this best be addressed?)

22
Some questions for discussion (Cont)

• What other questions should we discuss?