Scribd
Upload
41 views

Role and Participation of National and International Funding Agencies

This document discusses the role and participation of national and international funding agencies in developing exascale computing capabilities. It provides background on existing funding levels from various US agencies like DOE, NSF, and DARPA as well as European agencies. It emphasizes that exascale development requires an immediate, collaborative, and separate effort from current petascale work. Suggestions include developing joint roadmaps, increasing education/training programs, and establishing various collaboration models between funding agencies ranging from loose to tightly coupled partnerships. Clear definitions of agency responsibilities and tangible targets are needed to support this effort over the long term.

Uploaded by

Manoj Hegde
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
41 views

Role and Participation of National and International Funding Agencies

This document discusses the role and participation of national and international funding agencies in developing exascale computing capabilities. It provides background on existing funding levels from various US agencies like DOE, NSF, and DARPA as well as European agencies. It emphasizes that exascale development requires an immediate, collaborative, and separate effort from current petascale work. Suggestions include developing joint roadmaps, increasing education/training programs, and establishing various collaboration models between funding agencies ranging from loose to tightly coupled partnerships. Clear definitions of agency responsibilities and tangible targets are needed to support this effort over the long term.

Uploaded by

Manoj Hegde
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

Role and Participation of national and international funding agencies

Paul Messina, Fred Johnson, Bertrand Braunschweig, Catherine Riviere, Herve Mouren, Fabrizio Gagliardi, Abani Patra, Kostas Glinos

Background and Overview


Existing Funding US
DOE ($250M/year in s/w and algorithms for HPC)- OASCR $50M/year (SCIDAC fraction) computer science and applied mathematics both independently and in applications $60M in base programs NNSA ~$100M (30% Univ, 70% Lab) Retargetable $50M??

NSF $50M (diffuse across many programs),


OCI, CISE coordinate-- HECURA, PetaApps, CDI, SDCI, STCI, CISE/CCF) Retargetable $10M (guesstimate)

DARPA (homework!)

Background
Existing Funding Europe
ANR (National Agency for Research) Euro 25M -- in a mix of application and systems software GENCI -- substantial investment in PRACE Euro 25M national detailed agreements being worked out BMBF Euro 20M for HPC s/w Finland Euro 5M UK/Italy/Spain/NL/Others (homework!)

Background
EC (DEISA, PRACE investments, ) ~E40+M/yr
DEISA/PRACE etc. E20M/yr Computing Systems E12.5M/yr Tera-device Computing E7.5M/yr

Background
Japan ??? Other Asian???

Background
International Partnership Mechanisms
EC programs open to others -- developed countries pay their way;
ANR -- joint program possible but each country funds its own personnel; ANR target 7-8% international

Partnership is common but challenging-- time synchronization, Flexibility in NSF funding scenario using supplements NSF has Office of International Science and Engineering -- area specific expertise both at NSF and locally available

Motivation
Science & Engineering & Medicine & needs exascale (see before)! Challenges are increasingly GLOBAL (climate, energy )
Extreme computing now drives all science not just a few classified needs (e.g. ASCI)

Exascale will be different!


Existing stack not adequate/extensible Need ab initio effort Exascale effort needs to start ALONGSIDE but SEPARATELY from current tera/petascale efforts IMMEDIATELY Need Roadmap for planning actions and prioritizing investment areas

Risk Mitigation -- Collaborative efforts spread risk and cost Failure to coordinate can delay societal use of exascale But Challenges With Petascale will Persist
Need continuity of effort Peta --> exa Exascale communities will arise from existing petascale groups

Education & Training


Joint Education Training
Training needs to focus on exascale challenges of concurrency O(100) students and junior fellowships loosely affiliated to research at large scale computing in both US and EU; Computational Science Graduate Fellowship at DOE more focused but ONLY 10!! Extensive mobility among EU partners Need scaling up in numbers and partners, structured exchanges? Community formation?

Education and Training


Joint Education and Training
New programs in Germany and France HPC Masters -- 50 students/year in each, Teratec, Juelich HPC related PhD positions in France

Environment to develop Exascale


Weaknesses
Design of research program for the uncharted territory of the High Risk-High Return research needed now Industry does not perceive significant immediate ROI compared to other users of HPC; applications that are not at the table HPC needs careful quantification -- e.g. DOE workshop series for some areas of Science Application software vendors need to be involved ongoing efforts Funding for new technology adoption is not flexible enough for quick response needed
Availability of critical mass of technical personnel and resources

Environment to develop Exascale


Strengths
Science case for benefits from extreme scale computing in the grand challenges of society -- energy, life sciences etc. is clear HPC ROI over the last decades is significant -- recent National Academy study -- Reduction in product development cycle Availability of critical mass of technical personnel and resources at US national laboratories

Collaboration Scenarios
1. Almost no collaboration --Joint workshops only
Not adequate

2. Loosely Coupled Collaboration -Benchmarks defined


May be beneficial but inadequate in many cases Can be very beneficial in most areas Beneficial in some cases

3. Collaboration with standardization 4. Tightly Coupled Collaboration

Collaborations

Collaboration Elements
Joint roadmaps, periodic workshops Coordinated Investment Expertise driven joint projects Structured Framework
International Clearing house of what is funded Need policies that promote effective reuse Weakness in long term support and maintenance that needs to addressed in funding mechanisms

Open Source Model

Actions
Respond to Roadmap needs defined by the IESP/ DARPA (and other) groups Need new program? New components in existing programs? Need stable and longer term framework for collaborations
Existing Collaborations Bilateral and driven by individual initiative Need agreements at different levels -- laboratories, agencies, political? Need mechanisms for involving new stakeholders

Action
Clear definition of responsibilities for different groups? Top/down, Physics community model? Need more multilateral structured mechanisms? Funding agencies need identified and tangible targets
Programming models, compilers, frameworks Support for application drivers -- multiscale, multiphysics, uncertainty quantification

Timeline for Funding Agency Coordination


Meet in September to evaluate interim IESP plans - IESP workshop in Tsukuba in Oct 18-20 Draft plans by April 2010

You might also like