Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
The European Supercomputing
Research Programme
Dr Panagiotis Tsarchopoulos
Future and Emerging Technologies
DG CONNECT
European Commission
EU by the numbers
28
Over 508million
combined population
1.6million
Member
States
Official
languages
24
of global
GDP
24%
of
GLOBAL EXPORTS
& IMPORTS
15% Square miles total surface area
<1/2 of U.S. surface area
.
Horizon 2020 - the EU framework programme for research
and innovation 2014-2020: a stronger, clearer focus
Excellent Science 24,4 B€
Industrial Leadership
17B€
Societal Challenges
29,6 B€
H2020 Budget: 77B€
OTHERS: 5,8B€ (Spreading excellence & widening participation, Science & Society, JRC, EIT)
(current prices)
H2020 covers the full innovation chain
Basic Research
Technology R&D
Demonstration Market uptake
Large scale validation
Prototyping Pilots
Excellent science
Industrial leadership
Societal challenges
Public-Private Partnership with ETP4HPC (1st January 2014)
Key EU policy developments in HPC
Communication from the EC "High-Performance Computing: Europe's place in a global race" (2012)
Council Conclusions on High-Performance Computing (Competitiveness Council – 2013)
Establishment of the European Technology Platform on High-Performance Computing (ETP4HPC - 2012) and Strategic Research Agenda on HPC (2013)
Horizon 2020 programme including HPC Calls adopted (end of 2013)
HPC strategy combining three elements:
(a) Computer Science: towards exascale HPC; A special FET
initiative focussing on the next generations of exascale computing as a key horizontal enabler for advanced modelling, simulation and big-data applications [HPC in FET]
(b) achieving excellence in HPC applications; Centres of Excellence
for scientific/industrial HPC applications in (new) domains that are most important for Europe [e-infrastructures]
(c) providing access to the best supercomputing facilities and services for both industry and academia; PRACE - world-class HPC
infrastructure for the best research [e-infrastructures]
• complemented with training, education and skills development in HPC
An integrated HPC approach in H2020
"Excellent Science" part of H2020
Excellence in HPC applications (Centres of Excellence)
EU development of Exascale technologies
Access to best HPC for industry and academia
Interrelation between the three elements
• specifications of exascale prototypes
• technological options for future systems
• Collaboration of HPC Centres and application CoEs
• provision of HPC capabilities and expertise
• identify applications for co-design of exascale systems
• Innovative methods and algorithms for extreme parallelism of traditional/emerging applications
"Excellent Science" part of H2020
PRACE is an international not-for-profit association under Belgian law,
with its seat in Brussels.
PRACE counts 25 members and 2 observers.
The PRACE Hosting Members are France, Germany, Italy and Spain.
PRACE is governed by the PRACE Council in which each member
has a seat. The daily management of the association is delegated to
the Board of Directors.
PRACE is funded by its members as well as through a series of
implementation projects supported by the European Commission.
Partnership for Advanced Computing in Europe
The European HPC Ecosystem: Access to best HPC
www.prace-ri.eu
MareNostrum: IBM
BSC, Barcelona, Spain
JUQUEEN: IBM
BlueGene/Q
GAUSS/FZJ
Jülich, Germany
CURIE: Bull Bullx
GENCI/CEA
Bruyères-le-Châtel, France
SuperMUC: IBM
GAUSS/LRZ
Garching, Germany
FERMI: IBM BlueGene/Q
CINECA , Bologna, Italy
Hazel Hen: Cray
GAUSS/HLRS,
Stuttgart, Germany
4 Hosting Members offering core hours on
6 world-class machines
PRACE
Access through PRACE Peer Review Criterion: Scientific Excellence
The European HPC Ecosystem: Access to best HPC
ETP4HPC Association, the European Technology Platform for High Performance Computing: An industry-led forum founded by stakeholders of HPC technology
Open membership for organizations having HPC R&D based in Europe
70 organizations are currently members
www.etp4hpc.eu
The European HPC Ecosystem: Technology development
Through the Strategic Research Agenda, the ETP4HPC Association has identified research areas and topics to reach a stronger European HPC environment that can benefit Europe and the rest of the world.
The European HPC Ecosystem: Technology development
Horizon 2020 may be implemented through Public-Private Partnerships where all the partners concerned commit to support the development and implementation of R&I activities of strategic importance to the Union.
The Union enters a contractual agreement (cPPP) with private partners
Commitment for the duration of the Horizon 2020 programme
Eight cPPPs launched in January 2014: Factories of the Future (FoF), Energy-efficient Buildings (EeB),
European Green Vehicles Initiative (EGVI), Sustainable Process Industry (SPIRE),
Advanced 5G network (5G), Robotics, Photonics
High Performance Computing
• Public-Private Partnership with ETP4HPC
• Based on the ETP4HPC Strategic Research Agenda
• Started 1st January 2014
• 700 m€ EU funding for the period 2014-2020
Public-Private Partnerships in H2020
• To build a European world-class HPC technology value chain that is globally competitive - synergy between the three pillars of the HPC ecosystem (technology development, applications and computing infrastructure)
• To achieve a critical mass of convergent resources in order to increase the competitiveness of European HPC vendors and solutions
• To leverage the transformative power of HPC in order to boost European competitiveness in science and business
• To expand the HPC user base, especially SMEs, and to facilitate the participation of SMEs in the provision of competitive HPC technology solutions
• To develop a EU leadership and world-wide excellence in key application domains for industry, science and society
• provision of innovative solutions for grand societal challenges
• development of the future applications for the exascale computing generation
Public-Private Partnership in HPC
General objectives
FETHPC-1-2014: HPC Core Technologies, Programming Environments and Algorithms for Extreme Parallelism and Extreme Data Applications (Research and Innovation Actions) a) HPC core technologies and architectures (e.g. processors, memory,
interconnect and storage) and their optimal integration into HPC systems, platforms and prototypes
b) Programming methodologies, environments languages and tools: new programming models for extreme parallelism and extreme data applications
c) Application Programming Interfaces and system software for future extreme scale systems
d) New mathematical and algorithmic approaches for existing or emerging applications
FETHPC 2014 CALL Summary
FETHPC-2-2014: HPC Ecosystem Development (Coordination and Support Actions) a) Coordination of the HPC strategy: coordination of the activities of
stakeholders, development of Strategic Research Agenda, mapping and analysis of national and international R&I programmes, attracting young talents ,…
b) Excellence in High Performance Computing Systems: boosting European research excellence on the key challenges towards the next generations of high-performance computing systems; cutting across all levels.
Budget: ~100 m€
Specific Challenge: Establish a limited number of user-centred Centres of Excellence (CoE) in the application of HPC for addressing scientific, industrial or societal challenges CoEs may be: Thematic, Transversal, Challenge-driven, or a combination Proposals (Research and Innovation Actions) will address:
a) Pan-European services
b) Exascale R&D
c) Sustainability
d) New communities
e) Governance structure
EINFRA-5-2015 CoEs Summary
Centres of Excellence for computing applications
Budget: 40m€
Results of the 2014-2015 Calls
• Projects start: Autumn 2015
• Complete HPC R&D project portfolio covering all elements of HPC stack
• Multinational, multi-partner consortia involved in all projects
• Specific actions to further develop the European HPC ecosystem
• Centres of Excellence for key HPC applications
Technology development
Compute
• Highly integrated, high-performance, heterogeneous SoC based on chiplets
• 3D integration with silicon interposer
• Design of a SoC/SoP with exascale-class performance
• Software ecosystem for the future SoC. • Building on the previous Mont-Blanc projects www.montblanc-project.eu
• Reconfigurable accelerators • Novel hierarchical architecture
www.ecoscale.eu
• Coarse-grain reconfigurable architectures
• Optimized just-in time synthesis tools
www.extrahpc.eu
exanode.eu
Interconnect, Memory, Storage
Technology development
www.sagestorage.eu
• Next-generation multi-tiered object-based data storage system
• Hardware and software
www.nextgenio.eu
• Scalable I/O using non-volatile memory • Hardware and software
• High throughput, low latency connectivity, suitable for exascale with congestion mitigation, QoS guarantees, and resilience
• Hyper-density system integration with total immersion liquid cooling
www.exanest.eu
Data-intensive Real-time
Technology development
• Exploration of deeply heterogeneous architectures allowing dynamic adaptation to applications
• Add Predictability dimension for design of real-time HPC systems compared to the mainstream Power/Performance HPC design dimensions
GREENFLASH • Design and build a prototype for a Real-Time
Controller targeting the European Extremely Large Telescope Adaptive Optics instrumentation (39m diameter telescope to see first light in the early 2020s; largest telescope on Earth)
• Tackle the combination of high data transfer bandwidth, low latency and high throughput requirements
www.mango-project.eu
Programming Technology development
• DSL for expressing adaptivity and auto-tuning strategies • Performance and energy control using software knobs • Advance the state of the art in operating applications at the
maximum energy-efficient and thermally-safe point www.antarex-project.eu
Parallel programming model based on nested recursive parallelism • automated porting of applications • multiple objective optimisation
www.allscale.eu
Automatic optimization detecting application's dynamic behaviour and adapting system environment at run-time
READEX
www.readex.eu
www.intertwine-project.eu
• Interoperability between programming models/APIs • 7 key APIs: MPI, GASPI, OpenMP, OmpSs, StarPU, QUARK, PaRSEC
www.hpc-escape.eu
• Defining and encapsulating the fundamental algorithmic building blocks ("Weather & Climate Dwarfs") underlying weather and climate codes
• Software stack for weather and climate codes at exascale
Mathematics & Algorithms
Technology development
New exascale hyperbolic simulation engine based on communication-avoiding schemes
• Parallel Numerical Linear Algebra • Novel algorithms for extreme-scale systems • Algorithm-based fault tolerance
NLAFET
• Exascale Fluid Dynamics Simulations • Error control and adaptive mesh refinement • Solver efficiency • Heterogeneous modelling
• Multiscale computing at exascale based on generic computing patterns • Approach tested on codes from fusion, biomedicine, materials and
astrophysics
• Exascale Compound Activity Prediction Engine • Machine learning for extreme-scale computing and data • Specific application: predicting compound bioactivity for pharmaceuticals
www.compat-project.eu
exahype.eu
www.nlafet.eu
exaflow-project.eu
www.excape-h2020.eu
Centres of Excellence Weather and Climate
Improve efficiency and productivity of numerical weather and climate simulation on high-performance computing platforms by supporting the end-to-end workflow of global Earth system modelling in HPC environment.
• Scalability of models, tools and data management on state-of-the-art supercomputer systems
• Usability of models and tools throughout the European HPC eco-system, and
• Exploitability of the huge amount of resulting data.
ESiWACE leverages two established European networks: • the European Network for Earth System modelling (ENES) http://enes.org representing the European climate
modelling community and • the world leading European Centre for Medium-Range Weather Forecasts (ECMWF) http://www.ecmwf.int
www.esiwace.eu
Energy Oriented Centre of Excellence
Accelerate the European transition to a reliable and low carbon energy supply Four renewable energy pillars with a heavy reliance on numerical modelling:
• Weather forecast, • Materials, • Water • Fusion These four pillars are supported by a strong transversal multidisciplinary basis providing high-end expertise in applied mathematics and HPC.
Centres of Excellence Energy
• Building on the network of CECAM (Centre Européen de Calcul Atomique et Moléculaire) www.cecam.org it will create a distributed, sustainable centre for simulation and modelling at and across the atomic, molecular and continuum scales.
• Actions: 1. development, testing, maintenance, and dissemination of robust software modules targeted at end-user needs; 2. advanced training of current and future academic and industrial researchers able to exploit these capabilities; 3. multidisciplinary, coordinated, top-level applied consultancy to industrial end-users (both large multinationals and SMEs).
Centres of Excellence Simulation and Modelling
An e-infrastructure for software, training and consultancy in simulation and modelling
www.e-cam2020.eu
Covering structural and functional studies of the building blocks of living organisms - proteins, DNA, saccharides, membranes, solvents and small molecules like drug compounds.
• Efficiency and scalability of key codes for biomolecular research;
• Usability of ICT technologies for biomolecular researchers in academia and industry;
• Best practices and train end users in making good use of both software and e-Infrastructure.
Centres of Excellence Biomolecular
Centre of Excellence for Biomolecular Research
bioexcel.eu
• Brings together HPC, HPDA and GSS to facilitate computer-aided policy making
• Development of an HPC-based framework to generate customized synthetic populations for Global System Science applications (a synthetic population provides a model of a given population, e.g. humans, plants, animals, cars, buildings etc.)
• Analytical methodology based on sound theoretical background from algorithmic game theory and a state-of-the-art understanding of socio-cultural evolution.
• Multi-dimension simulations with the help of synthetic populations for green growth, pandemics, urbanization
Centres of Excellence Global Systems
www.coegss-project.eu
Centre of Excellence for Global Systems Science
• Integrate leading codes and make their results comparable
• Building on the NoMaD Repository www.nomad-repository.eu
Centres of Excellence Materials
www.max-center.eu
Materials design at the eXascale
www.nomad-coe.eu
Porting, scaling, and optimising material science application codes for peta- and exascale platforms
Computing applications of any sort, from a few hundred to many thousand processors
? Application Performance Audit
•Primary service
•Identify performance issues of code
! Application Performance Plan
•Follow-up on the audit service
•Identifies the root causes of the issues found and qualifies and quantifies approaches to address them
√ Proof-of-Concept
•Experiments and mock-up tests for codes
•Kernel extraction, parallelization, mini-apps experiments to show effect of proposed optimizations
Centres of Excellence Performance
Performance Optimisation and Productivity
www.pop-coe.eu
European HPC Ecosystem development
www.eurolab4hpc.eu
• Long-term research roadmap: Computing beyond exascale
• HPC curriculum development
• Improve HPC academia links with business and venture capital
• Key action supporting the European Commission in the
coordination, development and implementation of a common strategy for the European HPC Ecosystem
• Run by PRACE and ETP4HPC
• Technology and applications roadmaps towards exascale
• Develop international research collaborations
www.exdci.eu
HPC Ecosystem
EXDCI Eurolab-4-HPC
ExCAPE
GreenFLASH
READEX ESCAPE
INTERTWINE
ALLScale
ANTAREX
ExaFLOW
ComPat
NLAFET
ExaHYPE
NEXTGenIO
SAGE
ExaNEST
ExaNoDe
ECOSCALE
EXTRA
Mont-Blanc 3 Mango
The new European HPC research landscape
Project Start: Autumn 2015
BioExcel COEGSS EoCoE E-CAM ESiWACE MAX NOMAD PoP
Centres of Excellence
Biomolecular Global
systems Energy
Simulation
Modelling Weather
Climate Materials Materials
Performance
optimisation
Future HPC Calls 2016-2017
HPC-related calls Horizon 2020
••• 32
• WP2016-2017
Support to the Pan-European HPC infrastructure (eInfrastructure)
• Implementation of PRACE (PRACE-4IP) – 15 m€
• Procurement of innovative solutions in HPC (PPI) – 26 m€
• HPC e-infrastructure for the Human Brain Project – 25 m€
Support to exascale technologies development (FET)
• Co-design of HPC systems and applications – 41 m€
• Transition to exascale computing – 40 m€
• Exascale Ecosystem Development – 4 m€
Headline research challenges FETHPC 2016/2017 Calls
• Co-design of HPC systems and applications: bigger projects with 10-20m indicative funding per project [2016]
• Transition to exascale computing : smaller focused projects with 2-4m indicative funding per project [2017] • High productivity programming environments for exascale
• Exascale system software and management
• Exascale I/O and storage in the presence of multiple tiers of data storage:
• Supercomputing for Extreme Data and emerging HPC use modes
• Mathematics and algorithms for extreme scale HPC systems and applications working with extreme data
• Exascale Ecosystem Development : Coordination and Support Actions with 1-2m indicative funding per project [2017]
HPC Overall strategy Horizon 2020 Calls 2014-2017
Applications Societal challenges
Scientific strategic applications Emerging domains (Big Data)
New methods and algorithms
Exascale technologies Architectures, programming.
environments, tools… Exascale Prototypes
Pan-European HPC Infrastructure
HPC Capability HPC Services
Support to innovation
Training Education
Skills
CLOUDS
ADVANCED COMPUTING
SMEs Services, Competence
Centres
Flagship Applications
(HBP)
Clouds for Science
PRACE-4IP (15 m€)
Centres of Excellence (40 m€)
Network of SME competence centres
(2 m€)
Core technologies (93,4 m€) Ecosystem development (4 m€)
HIGH PERFORMANCE COMPUTING
WP2014-15 ~155 M€ (~140 in the cPPP)
WP2016-17 ~151 M€ (85 in the cPPP)
PRACE (15 m€) PPI for HPC (26 m€)
HBP – HPC (25 m€)
Co-design (41 m€) Transition to exascale (40 m€) Ecosystem development (4 m€)
2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
l l l l l l l l l l
Projects from 2014 Call: Technology building blocks
& 1st round CoEs
Projects from 2016 Call: 1st level co-design
Projects from 2017 Call: Exascale transition
Future 2018/19/20 Calls: Extreme Scale Demonstrators
(2nd-level co-design / integration) &
2nd round of CoEs
HPC PPP timeline in H2020 (indicative)
HPC Ecosystem development
Projects start
Exascale Technologies & Applications
Examples of International Cooperation in HPC
• Wind energy production and design, • Efficient combustion systems for biomass-
derived fuels (biogas), • Exploration geophysics for hydrocarbon
reservoirs • EU-Brazil cooperation: 13 partners, 7 from
Europe and 6 from Brazil
www.hpc4e.eu
RISC (2011-2013): deepen strategic R&D cooperation between Europe (EU) and Latin America (LA) in the field of High Performance Computing (HPC).
HPC4E: exascale HPC for Energy
"Notre ambition c'est que d'ici 2020, l'Europe se
classe dans le top 3 mondial du calcul à haute
performance."
European Commission President
Jean-Claude Juncker
"Our ambition is that until 2020, Europe is in the world top 3 of high performance computing."
27 October 2015
Thank you for your attention!
Email: [email protected]
Single point of access to all information about the EC HPC strategy, work programmes and HPC related news:
ec.europa.eu/horizon2020-hpc
exdci.eu/events/european-hpc-summit-week