european exascale software...
TRANSCRIPT
FP7 Support Action - European Exascale Software Initiative
DG Information Society and the unit e-Infrastructures
European Exascale Software Initiative
SPXXL/SCICOMP, Paris - May 2011
Jean-Yves Berthou, EDF R&D
EESI coordinator, www.eesi-project.eu
1
Factors that Necessitate Redesign of Our Software
� Steepness of the ascent from terascale to petascale to exascale
� Extreme parallelism and hybrid design
� Preparing for million/billion way parallelism
� Tightening memory/bandwidth bottleneck
� Limits on power/clock speed implication on multicore
� Reducing communication will become much more intense
� Memory per core changes, byte-to-flop ratio will change
� Necessary Fault Tolerance
� MTTF will drop� Checkpoint/restart has limitations
Software infrastructure does not exist today
Potential System Architecture Targets
A Call to Action� Hardware has changed dramatically while software
ecosystem has remained stagnant� Previous approaches have not looked at co-design of
multiple levels in the system software stack (OS, runtime, compiler, libraries, application frameworks)
� Need to exploit new hardware trends (e.g., manycore, heterogeneity) that cannot be handled by existing software stack, memory per socket trends
� Emerging software technologies exist, but have not been fully integrated with system software, e.g., UPC, Cilk, CUDA, HPCS
� Community codes unprepared for sea change in architectures
� No global evaluation of key missing components
www.exascale.org
International Community Effort
� We believe this needs to be an international collaboration for various reasons including:� The scale of investment� The need for international input on requirements � US, Europeans, Asians, and others are working on their
own software that should be part of a larger vision for HPC.� No global evaluation of key missing components� Hardware features are uncoordinated with software
development
www.exascale.org
Where We Are Today:Where We Are Today:
www.exascale.org
� Ken Kennedy – Petascale Software Project (2006)� SC08 (Austin TX) meeting to generate interest
� Funding from DOE’s Office of Science & NSF Office of Cyberinfratructure and sponsorship by Europeans and Asians
� US meeting (Santa Fe, NM) April 6-8, 2009 � 65 people
� European meeting (Paris, France) June 28-29, 2009
� Outline Report� Asian meeting (Tsukuba Japan) October 18-20, 2009
� Draft roadmap and refine report
� SC09 (Portland OR) BOF to inform others
� Public Comment; Draft Report presented� European meeting (Oxford, UK) April 13-14, 2010
� Refine and prioritize roadmap; look at management models
� Maui Meeting October 18-19, 2010
� SC10 (New Orleans) BOF to inform others (Wed 5:30, Room 389)
� San Francisco Meeting – April 6-7, 2011
Apr 2009
Jun 2009
Oct 2009
Nov 2009
Apr 2010
Oct 2010
Nov 2008
Nov 2010
Apr 2011
IESP Goal
Build an international plan for developing the next generation open source software
for scientific high-performance computing
Build an international plan for developing the next generation open source software
for scientific high-performance computing
Improve the world’s simulation and modeling capability by improving the coordination and development of the HPC software environment
Workshops:
www.exascale.org
Roadmap Components
www.exascale.org
Motivations for launching EESI
Coordinate the European contribution to IESP
Enlarge the European community involved in the software roadmapping activity
Build and consolidate a vision and roadmap at the European Level, including applications, both from academia and industry
11/05/2011 9
Characteristics of the EESI project
Coordination and support action – FP7/Infrastructureswww.eesi-project.eu
Coordinator: EDF R&D, Jean-Yves Berthou Starting date : 1st of June 2010, for 18 monthsRequested EC contribution : 640 000 €Consortium : 8 contractual partners:
17 associated participants11 contributing participants
The project has received funding from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 261513 1010
EESI partners
11
-
STRATOSSTRATOSSTRATOSSTRATOS
11
EESI main goals
Build a European vision and roadmap to address the challenge of performing scientific computing on the new generation of computers which will provide hundreds of Petaflopperformances in 2015 and Exaflop performances in 2020
� investigate how Europe is located, its strengths and weaknesses, in the overall international HPC landscape and competition
� identify priority actions� identify the sources of competitiveness for Europe induced by the
development of Peta/Exascale solutions and usages� investigate and propose programs in education and training for the
next generation of computational scientists� identify and stimulate opportunities of worldwide collaborations
1212
EESI main tasks
Coordination of the European participation in IESP� Make a thorough assessment of needs, issues and strategies
� Develop a coordinated software roadmap� Provide a framework for organizing the software research community
� Engage and coordinate vendor community in crosscutting efforts
� Encourage and facilitate collaboration in education and training
Cartography of existing HPC projects and initiatives in Europe, US and ASIA
Coordination of “disciplinary working groups” at the European level� Four groups “Application Grand Challenges”
� Four groups “Enabling technologies for Petaflop/Exaflop computing”
Synthesis, dissemination and recommendation to the European Commission
1313
Initial cartography of existing HPC projects, initiatives in Europe, US and ASIA
14
� Initiatives in Europe:� Europe – Funding
� Europe – Collaborative Projects & Existing Initiatives
� National Initiatives in Europe
� Initiatives worldwide:� US Systems & Initiatives
� Japan Systems & Initiatives� China Systems & Initiatives
� South Korea Systems & Initiatives
� India Systems & Initiatives
www.eesi-project.eu
15
5 months 8 months 3 monthsT0+5 T0+17 T0+18T0+13
1 monthT0
1 monthT0+12
1 monthT0+16
June 1, 2010 October 2010 June 2011 November 2011
Initial International
workshop
Internal workshop: presentation of each working group results
and roadmaps
Synthesis of all contributions and
production of a set of
recommendations
Initial cartography of existing HPC
projects, initiatives in Europe, US and
ASIA
Weather, Climatology and Earth Sciences
Industrial and Engineering Applications (Transport, Energy)
Life science, Health, BPM
Fundamental Sciences (Chemistry, Physics)
Scientific software engineering
Software eco-systems
Hardware roadmap, links with vendors
Numerical libraries, solvers and algorithms
Updated cartography of existing HPC
projects, initiatives in Europe, US and
ASIAT0
European Exascale Software Initiative AGENDA
Constitution of WG, Setup of guidelines, organisation modes
Enabling technologies for Exaflop computing
Application Grand Challenges
Presentation of EESI
results to the
EC
Final conference:
public presentation of project result
15
� EESI General framework
WP3 and WP4 Interactions
16
WP3 – Applications Grand Challenge
WP4 – Enabling Technologies for EFlops Computing
Scientific needs
Hardware & softwareroadmaps
WP2 : Link with IESP and other projects
WP5 : Communication - Dissemination
IESP Meeting, San Francisco, 6-7 April 2011
� Common templates in order to
� Facilitate exchanges between WP3 and WP4� Identify and classify key issues (What-Who-Where-When-How much )
� Description of the scientific and technical perimeter of the WG (*)
� Social benefits, societal, environmental and economical impact (*)
� Scientific and technical hurdles� Address cross cutting issues : Resilience, Power Mngt, Programmability,
Performance and Reproducibility of the results, …� European strengths and weaknesses in the worldwide competition
� Sources of competitiveness for Europe
� Needs of education and training� Potential collaborations outside Europe
� Existing funded projects and funding agencies
� Timeline, needs of HR, provisional costs, ...
� Building an (or several) exa-scale prototype in Europe? By when?
� Facilitate elaboration of intermediate and final reports(*) for Application WGs, description in terms of Application Grand Challenges
Common Guidelines for Group Activity
17 IESP Meeting, San Francisco, 6-7 April 2011
� 115 experts excluding Chairs/Vice chairs� 14.4 experts/group av without Chairs/Vice Chairs� Still possible to enrol up to 5 additional experts
� 14 countries, good European coverage� Participation of 2 US, 1 Israelian and 1 Lithuanian
experts
Expert distribution
18EESI WorkshopNovember 9 2010
EESI participants
1919
-
� Enabling technologies for Exaflop computing (JSC)� Assess novel HW and SW technologies for Exascale challenges� Review IESP roadmap and discuss in broader EU context� Economic impact & European competitiveness� Build a European vision and a roadmap
� 4 working groups� WG 4.1 : Hardware roadmap, links with Vendors
H. Huber (Chair, STRATOS/LRZ) & R. Brunino (Vice Chair, CINECA)� WG 4.2 : Software eco-systems
F. Cappello (Chair, INRIA) & B. Mohr (Vice Chair, JSC)� WG 4.3 : Numerical libraries, solvers, and algorithms
I. Duff (Chair, SFTC) & A. Grothey (Vice chair, Univ. Edinburgh)� WG 4.4 : Scientific software engineering
M. Ashworth (Chair, STFC) & A. Jones (Vice Chair, NAG)
WP4 Enabling technologies for Exaflop computing Organisation
20 IESP Meeting, San Francisco, 6-7 April 2011
WG 4.1 : Hardware roadmap, links with Vendors
21
Leif Nordlund AMD SE Business Development Manager
Jean-Pierre Panziera Bull FR Director of Performance Engineering
Wilfried Oed Cray DE HPC Engineer, EMEA
Luigi Brochard IBM FR Distinguished HPC Engineer
Andrey Semin Intel RUS HPC Technology Manager, EMEA
Frank Baetke HP DE Worldwide Director of High Performance Computing
Michael Kagan Mellanox IL CTO of Mellanox Technologies
Dev Tyagi Supermicro UK General Manager at Supermicro, UK
Giampetro Tecchiolli Eurotech IT CEO/CTO of Eurotech
Salvatore Rinaudo ST Microelectronics IT ST Microelectronics
Rüdiger Wolff SGI DE
Principal Systems Engineer & Systems Engineer Manager, EMEA
Dmitry Tkachev T-Platforms RUS Chief Architect
Pierre Lagier Fujitsu FR Technical Director
Ulrich Brüning ExTOLL DE Computer Architecture Research Manager
Aad van der Steen NCF NL HPC architectures & benchmarking
Alex Ramirez BSC ES Computer Architecture Research Manager
Dominik Ulmer CSCS CH HPC Center
Jean Gonnord CEA FR Numerical Simulation & Computer Sciences
Questions for the experts� What are the major challenges/issues in HPC in the next 5 to 10 years?� What efforts will your institution or company under take to address these
challenges/issues?� a) If applicable, please mention all R&D on promisi ng future multi-Peta to
Exascale hardware technologies such as hybrid many-core CPUs, novel memory technologies (3D-stacked memory, memristor, Spin Torque Transfer Magnetic RAM, graphene based memory, etc.), novel background storage technologies, photonic interconnects, highl y cooling efficient system packaging technologies, etc.b) If applicable, please mention all R&D on promisi ng future multi-Peta to Exascale software technologies such as highly scalable programming languages, fault tolerant parallel programming envi ronments, parallel file systems with end-to-end data integrity, highly scal able and energy efficient system management software, hierarchical storage so lutions, OS with coherent inter-node scheduling mechanisms, etc.c) If applicable, please mention all R&D on promising highly scalable numerical algorithm for multi-Peta to Exascale applications
22
WG 4.1 : Hardware roadmap, links with Vendors
Chairs:� Franck Cappello INRIA&UIUC, 1 FR, Resilience and FT� Benrd Mohr, JSC, 1 DE, Performance tools
� Jesus Labarta BSC 1 ES Programming models� Marc Bull EPCC 1 UK OpenMP / PGAS
� François Bodin CAPS 2/V FR Programming/Compiler for GPUs� Raymond Namyst LABRI Bordeaux, 3 FR MPI&OpenMP Runtime , GPUs
� Jean-François Méhaut INRIA Grenoble 4 FR Performance Modeling/Apps.
� Matthias Müller TU Dresden 2 DE Validation/correctness Checking � Felix Wolf GRS 3 DE Performance Tools� David Lecomber ALINEA 2/V UK Parallel debugger � Simon McIntosh-Smith U. Bristol 3 UK Computer Architecture & FT
� Vladimir Voevodin MSU 1 RU Performance tools � Thomas Ludwig DKRZ 4 DE Power management � Olivier Richard INRIA 5 FR Job and Resource Manager� Jacques C. Lafoucriere CEA 6 FR I/O, File system � Toni Cortés BSC 2 ES I/O, Storage
� Pascale Rossé BULL V FR OS, System management� Karl. Solchenbach Intel V BE All
WG 4.2 : Software eco-systems
23
IESP San Francisco, April 2011
WG 4.2 TOPICS� System Software
� Operating Systems, System Management� Job and Resource Manager � Runtime Systems � I/O systems
� Development Environments� Programming Models, Compilers � Debuggers� Performance tools� Correctness tools
� Crosscutting Dimensions� Resilience� Power management
WG 4.2 : Software eco-systems
IESP San Francisco, April 2011
WG 4.3 Numerical Libraries, Solvers and Algorithms
25
Expert Name Field of Expertise Organisation Country
Jack DongarraHPC. Numerical linear algebra
Manchester/Tennessee UK/USA
Mike Giles GPU. CFD/Finance Oxford University UK
Gerard Meurant HPC. PDE solution ex-CEA Paris FR
Volker MehrmannLinear algebra. HPC applications
TU Berlin/Matheon DE
Torsten Koch Combinatorial Optimization BerlinDE
Peter Arbenz Eigenvalues. HPC ETH Zurich CH
Bo Kågström Dense Linear Algebra Ulmea SE
Julius Žilinskas Global optimization Vilnius University LT
Salvatore Filippone HPC. Numerical software Tor Vergata, Rome IT
Luc Giraud Iterative and hybrid methods INRIA Bordeaux FR
Patrick Amestoy Direct methods. Solvers ENSEEIHT-IRIT, Toulouse FR
Karl Meerbergen Preconditioners KU Leuven BE
Francois Pellegrini Mesh, load balancingBordeaux University, INRIA lab LABRI
FR
� Chair:
Iain
Duff
(STFC)
� Vice-Chair:
Andreas
Grothey
(EPCC)
IESP Meeting, San Francisco, 6-7 April 2011
WG 4.4 Scientific Software Engineering
26
Expert Name Field of Expertise Organisation Country
Sebastian von Alfthan HPC algorithms and optimisation CSC FI
John Biddescombe scientific visualisation CSCS CH
Ian Bush HPC algorithms and optimisation NAG UK
Iris Christalder HPC software frameworks LRZ DE
Rupert Ford HPC algorithms and optimisationUniversity of Manchester
UK
Yvan Fournier HPC algorithms and optimisation EDF FR
Joachim Hein HPC algorithms and optimisation Lund University SE
Mohammed Jowkar HPC algorithms and optimisation BSC ES
Peter Michielse HPC algorithms and optimisation NWO NL
Stephen Pickles HPC algorithms and optimisation STFC UK
Felix Schuermann Blue Brain project EPFL CH
Stephane Ploix Visualisation EDF FR
Christopher Greenough Software engineering STFC UK
Ash Vadgama Performance Modeling AWE UK
� Chair:
Mike
Ashworth
(STFC)
� Vice-Chair:
Andrew
Jones
(NAG)
IESP Meeting, San Francisco, 6-7 April 2011
� Objectives� Identify issues for Exascale application development� Develop roadmap for an Exascale application development
process
� First results� Five themes were proposed and accepted:
� Application frameworks and workflows� Visualization and data management� Fault tolerant algorithms� Application design� Software engineering
� Each themes was investigated and extensively discussed� Cross-cutting issues were identified
WG 4.4 Scientific Software Engineering
27 IESP Meeting, San Francisco, 6-7 April 2011
� Application Grand Challenges (GENCI)� Identify the apps drivers for Peta and Exascale� Needs & expectations of scientific applications� Economic impact & European competitiveness� Build a European vision and a roadmap
� 4 working groups� WG 3.1 : Industrial & Engineering Apps (Transport, Energy)
P. Ricoux (Chair, Total) & JC. André (Vice Chair, CERFACS)� WG 3.2 : Weather, Climatology and Earth Sciences
G. Aloisio (Chair, ENES-CMCC) & M. Cocco (Vice Chair, INGV)� WG 3.3 : Fundamental Sciences
G. Sutmann (Chair, CECAM-JSC) & JP. Nominé (Vice chair, CEA)� WG 3.4 : Life Sciences and Health
M. Orozco (Chair, BSC) & J. Thorton (Vice Chair, EBI)
WP3 Application Grand Challenges Organisation
28 IESP Meeting, San Francisco, 6-7 April 2011
� Aeronautics: full MDO, CFD-based noise simulation real-time CFD-based in-flight simulation: the digital aircraft
� Structure calculation: design new composite compounds, …� Special Chemistry : atom to continuum simulation for macro
parameters estimation in catalyst, surfactants, tribology, interfaces, nano-systems, …
� Energy: computations with smaller and smaller scales in larger and larger geometries : Turbulent combustion in closed engines and opened furnaces, explosion prediction, nuclear plants, hydraulics, …
� Oil and gas industries: full 3D inverse waveform problem, multi-scale reservoir simulation models, …
� Engineering (in general): multi-scale CFD, multi-fluids flows, multi-physics modelling,, complex systems modelling, …
WG 3.1 : Industrial & Engineering Apps (Transport, Energy)
29
Lead by: P. Ricoux (Chair, Total) & JC. André (Vice Chair, CERFACS)
Geophysics, Oil and gasMultiphase flows, Oil and gasCombustion, CFDFlight physics, Aeronautics
Applied math., PhysicsHPC, Particle dynamicsCFD, HydraulicsPropulsion and engine flows, AutomotivePropulsion, Engine flowsCFD, AeronauticsNeutronics, Nuclear industrySurface transportation, Trains Chemical EngineeringComputingComputer Aid Engineering
WG 3.1 : Industrial & Engineering Apps (Transport, Energy)
30
3-55 Hz
Code_Saturne-THYC
Lead by: P. Ricoux (Chair, Total) & JC. André (Vice Chair, CERFACS)
Henri CALENDRA (TOTAL, F)Keld NIELSEN (ENI, I) Thierry POINSOT (CERFACS, F)Eric CHAPUT (EADS/Airbus, F)Demetrios PAPAGEORGIOU (Imp. College, GB)Ulrich RUDE (U. Erlangen, D)Jean-Daniel MATTEI (EDF, F)Christian HASSE (BMW, D)Heinz PITSCH (Univ. Aachen, D)Norbert KROLL (DLR, D) Tanguy COURAU (EDF, F)Ali TABBAL (ALSTOM, F)Chuan-yu WU (U. Birmingham, GB)Mark SAWYER (Univ. Edimburgh)Jacek MARCZYK Ontonix
Questions and issues to address:- for efficiently reaching goals: which effort ? where to spend the money? - the roadmap must be implemented in several industriesRoadmaps are different depending upon particular indu stries : For some domains, Exaflop capacities are needed to solve industrial
problems:Aeronautics: Exaflop not the ultimate goal, need at least Zetaflop for LES
aircraft. Many “farming” applications (almost independent simulations), Data mining and data assimilation, Steady cases (RANS, URANS) with improved physical modeling
Seismics, O&G: Largely embarrassingly parallel, major problems are programming model, memory access and data management, need Zetaflop for full inverse problem
Engineering: Optimization, Monte Carlo type simulations, LES/RANS/DNS … (main problem : equations themselves, physics, coupling, …)
For these domains, production problems will be solved by “farming”
applications . 31
WG 3.1 : Industrial & Engineering Apps (Transport, Energy)
For some domains, Exaflop capabilities are needed to solve industrial problems:
Combustion: turbulent combustion modeling, LES in large scale reactors, Coupling multi-physics, Exaflop for Combustion at the right scale, Multi-cycle engine ( Weak scalability)
Nuclear Energy: steady CFD calculations on complex geometries , RANS, LES and quasi-DNS type calculations under uncertainties, Monte Carlo Ultimate Neutronic Transport Calculation (Time dependant & Multiphycis coupling)
Other domains: multi-Fluids Flows , Fluid Structure Interactions, Fluidized beds (particle simulations , stochastic PDE, Multi-scale approach, physics, fluid in particles
32
WG 3.1 : Industrial & Engineering Apps (Transport, Energy)
Common main issues to be addressed (1/2)
� At the level of the simulation environment:� Unified Simulation Framework and associated services: CAD, mesh generation, data setting tools, computational scheme editing aids, visualization, etc.
� Multi‐physics simulation: establishment of standard coupling interfaces and software tools, mixing legacy and new generation codes
� Open‐source jointly developed mesh‐generation tool, automatic and adaptive meshing
� Standardized efficient parallel IO and data management (sorting memory for fast access, allocating new memory as needed in smaller chunks, treat parts of memory that are rarely/never needed based on heuristic algorithms, …)
33
WG 3.1 : Industrial & Engineering Apps (Transport, Energy)
Common main issues to be addressed (2/2)
� At the level of codes/applications:� New numerical methods, algorithms, solvers/libraries, improved efficiency
� coupling between stochastic and deterministic methods
� Numerical scheme involving Stochastic HPC computing for uncertainty and risk quantification
� meshless methods and particle simulation
� Scalable program, strong and weak scalability, load balancing, fault-tolerance techniques, multi-level parallelism, issues identified with multi-core with reduced memory bandwidth per core , collective communications, efficient parallel IO
� Standards programming models (MPI, OpenMP, C++, Fortran, …) handling multi-level parallelism and heterogeneous architecture
� Human resources, training (what level?)34
WG 3.1 : Industrial & Engineering Apps (Transport, Energy)
The big challenge is to model this complex system
Warren M. Washington – NCAR
Scientific Grand Challenges Workshop Series:
Challenges in Climate Change Science and the Role of Computing at the Extreme Scale
DOE Workshop (ASCR-BER)
November 6-7, 2008
• Several complex process to be simulated
• Several interacting processes
• Great range of time scales to be
analyzed
• Great range of spatial scales to be
considered
• Need interdisciplinar sciences (physics,
chemistry, biology, geology,…)
• Inherently non-linear governing
equations
• Need sophisticated numerics
• Need huge computational resources
WG 3.2 : Weather, Climatology and Earth Sciences
Lead by : G. Aloisio (Chair, ENES-CMCC) & M. Cocco (Vice Chair, INGV)
36
WG 3.2 : Weather, Climatology and Earth Sciences
DE 3FR 3IT 3UK 3SE, FI, ES 3
Organization Email Country Area of Expertise
ENES-CMCC [email protected] IT Exascale Computing ChairINGV [email protected] IT Seismology Vice-ChairCMCC [email protected] IT Oceanography
IPSL [email protected] FR Earth-system modellingCERFACS [email protected] FR Coupled Climate ModelsIPG [email protected] FR Solid Earth, Seismology
BADC [email protected] UK Climate Data ManagementManchester Univ. [email protected] UK High Performance Computing ECMWF [email protected] UK Data Assimilation
MPI [email protected] DE Earth System ModelingDKRZ [email protected] DE High Performance Computing LMU [email protected] Solid Earth, Geophysics
SMHI [email protected] SE Climate Change modelingFMI [email protected] FI MeteorologyBSC [email protected] ES Earth Sciences
(Final) Topics Group
Lead by : G. Aloisio (Chair, ENES-CMCC) & M. Cocco (Vice Chair, INGV)
37
WG 3.2 : Weather, Climatology and Earth Sciences
The WCES (Weather, Climate & Earth Sciences) Main Goals
� High-resolution Climate and Earth System Models to better simulate the interactions and feedbacks among the physical & biological complex processes controlling natural phenomena
� Innovative approaches for modelling solid Earth processes to better understand the physical processes controlling earthquakes, volcanic eruptions and tsunamis as well as those driving tectonics and Earth surface dynamics
• atmosphere• ocean• land, soils, permafrost, and vegetation (specified and interactive) cover• sea ice• land ice• carbon and other biogeochemical cycle• clouds and related microphysics• hydrology• atmospheric chemistry• aerosols• ice sheets• human systems
Several complex components to be considered !!!
� Unified next-generation atmospheric simulation systems for Weather and Climate scientists, allowing a strong reduction in climate prediction uncertainty
Lead by : G. Aloisio (Chair, ENES-CMCC) & M. Cocco (Vice Chair, INGV)
ENES foresight document (working draft)
Evolution diagram with global/regional models and resolution
Ensemble runs are needed for quantifying uncertainty
WG 3.2 : Weather, Climatology and Earth Sciences
Critical Exascale Software for WCES
Multiscale
WG 3.3: Fundamental Sciences
40
Particles
Plasmas
Atoms
Molecules
Time and space scales
…
Large structures
Domain
sQCD
Fusion for energy
Astrophysics
Nanosciences
Laser/plasmaInteraction
Chemistry/catalysis
Soft matter/polymers
Quantum chemistryElectronic structure
Lead by: G. Sutmann (Chair, CECAM-JSC) & JP. Nominé (Vice chair, CEA)
Fundamental sciences: Physics, Chemistry, Material Sciences, Astrophysics
Name Organization Country Area of ExpertiseVolker Springel Garching, MPI Astrophysik Ger AstrophysicsRomain Teyssier ETH Zürich Sui AstrophysicsMaurizio Ottaviani CEA Fra Fusion
Luis SilvaUniversidade Tecnica de Lisboa Por Laser Plasma Interaction
Alexei Removich Kokhlov Moscow State University Rus Soft MatterAlessandro Curioni IBM Research - Zurich Sui Materials SciencesGilles Zerah CECAM - CEA Fra Materials SciencesNicola Marzari University of Oxford UK Materials SciencesAdrian Wander STFC Daresbury UK Materials Sciences
Mike Payne University of Cambridge UK Quantum ChemistryThierry Deutsch CEA Fra Quantum ChemistryMike Ashworth STFC Daresbury UK Methods and algorithmsThomas Schultess CSCS Sui Materials SciencesPieter in t'Veld BASF Ger Soft matter
41
Ger: 2 Sui: 3 Fra: 3 Por: 1 Rus: 1 UK: 4 Total: 14
WG 3.3: Fundamental Sciences
� Objectives (1)� Identify exascale challenging problems like
� Quantum chemistry: electronic structure calculations for large systems, excited states
� Materials Science: ceramics, semi-conductors, design of nano-particles, catalysis
� Soft matter physics: simulation of long polymers, self organisation
� Astrophysics: cosmology, high-resolution solar physics
� Plasma physics: fusion and laser-plasma interactions � Particle physics
What is the impact of: fundamental understanding, industrial impact
WG 3.3: Fundamental Sciences
42
Science Drivers
43
� Materials Science : first principles design of materials� e.g. energy conversion
• chemical energy to electricity (fuel cells)
• sunlight to electricity (photovoltaic cells)
• nuclear energy to electricity (nuclear fission or fusion)
� e.g. energy storage• batteries
• supercapacitors
• chemical storage in high energy density fuels (hydrogen, etanol, methanol)
� e.g. energy use• solid state lighting, smart windows, low power computing, leightweightmaterials for transporting
WG 3.3: Fundamental Sciences
Summary
44
� Software issues for Fundamental Sciences
� Fault tolerant and energy efficient algorithms� Software support to measure or estimate energy
� Data locality
� Optimal order algorithms
� Mesh generation
� Algorithms with low communications� Standard interfaces for multiscale simulations
� Support of several executables in one job� Parallel I/O
� Load-balancing
WG 3.3: Fundamental Sciences
Lead by: M. Orozco (Chair, BSC) & J. Thorton (Vice Chair, EBI)
Simulation scenario in Life Sciences� Molecular simulations
� Structural prediction� Docking� Atomistic simulation� Cell-scale
mesoscopicsimulations
� Gene inter-relations� Cell simulation� Organ simulation� Ecosystem simulation� Drug design
J.M.Cela
WG 3.4 : Life Sciences and Health
46
Helmut Grubmuller Max Planck InstitutePaolo Carloni German Research SchoolWofgang Wenzel Karlsruhe Institute of TechnologyReinhard Schneider EMBL
Janet Thorton/Andrew Lyall EMBL-EBICharles Laughton University of Nottingham
Modesto Orozco BSCAlfonso Valencia CNIO
Richard Lavery University of NottinghamNicolas Baurin Sanofi-Aventis
Erik Lindahl Stockholm University
Henry Markram/Felix Schuermann EPFLManuel Peitsch Swiss Institute of Bioinformatics
Anna Tramontano University of RomaJulian.Tirado-Rives (US) University of Yale Adrian E. Roitberg (US) University of Florida
WG 3.4 : Life Sciences and Health
Example of MD Simulations of biological systems :
� 1000 times bigger, realistic cell membrane models with drug permeation and binding => large amount of data
� 1000 times more computationally complex, simulations of biomolecules based purely on quantum mechanical principles , possible at Exascale but need high memory per core
� Timescales 1000 times longer, studies of the dynamics of protein folding and molecular motors that take up t o seconds of real time , Anton (a specific purpose architecture) is capable to run 100 longer (3~5 years ahead)
WG 3.4 : Life Sciences and Health
Felix Schürmann, Herny Markram (Blue Brain Project, EPFL)
Medical simulation: Human Brain Project
WG 3.4 : Life Sciences and Health
Summary� Molecular Dynamics
� Data expected to grow 10~100 times, competitiveness vs ASIC, Exascale capability for complex systems and biased-sampling techniques
� Genomics � Data management, high I/O requirement, constant application
development, training
� Systems Biology � Multi-scale simulation, visualization tools, hardware flexibility for
new software development, data handling
� Medical Simulation� Data Management, real-time monitoring, Exabyte memory, data
transfer, Multi-scale
WG 3.4 : Life Sciences and Health
50
5 months 8 months 3 monthsT0+5 T0+17 T0+18T0+13
1 monthT0
1 monthT0+12
1 monthT0+16
June 1, 2010 October 2010 June 2011 November 2011
Initial International
workshop
Internal workshop: presentation of each working group results
and roadmaps
Synthesis of all contributions and
production of a set of
recommendations
Initial cartography of existing HPC
projects, initiatives in Europe, US and
ASIA
Weather, Climatology and Earth Sciences
Industrial and Engineering Applications (Transport, Energy)
Life science, Health, BPM
Fundamental Sciences (Chemistry, Physics)
Scientific software engineering
Software eco-systems
Hardware roadmap, links with vendors
Numerical libraries, solvers and algorithms
Updated cartography of existing HPC
projects, initiatives in Europe, US and
ASIAT0
European Exascale Software Initiative AGENDA
Constitution of WG, Setup of guidelines, organisation modes
Enabling technologies for Exaflop computing
Application Grand Challenges
Presentation of EESI
results to the
EC
Final conference:
public presentation of project result
50
EESI Final Conference, www.eesi-project.eu
51
EESI Final InternationalConference10 – 11 October 2011,
World Trade Center,Barcelona (Spain)
ObjectivesThe main objective of the EESI Final International Conference is presenting the project results to the stakeholders, scientists and policy makers containing presentations and discussions with all experts participating to the WG activities.
Policy makers, scientists, and participants of the EESI project are invited to attend this event.