CCFE is the fusion research arm of the United Kingdom Atomic Energy Authority.
This work was part-funded by the RCUK Energy Programme [grant number EP/I501045]
and the European Union’s Horizon 2020 research and innovation programme.
Small Scale HPC for Big Science
Shaun de Witt
shaun.de-witt@ukaea.uk
EOSC-Hub Public Day
Malaga, April 2018
• Background
• EOSCpilot Work and Preliminary Results
• EOSC-Hub Planned Work
• Conclusions
Agenda
• Making a Sun on Earth
– Fusing nuclei together
– Brings together compute intensive disciplines
• UQ, FEM, Radiative Transfer, Neutronics, spin lattice
dynamics,
• The Numbers – JET/ITER:
– Plasma Temperature: >100x106 (150x106) C
– Plasma Density: ~10-6 (10-5) atmospheric
– Duration of ‘shot’: 5-20 sec (350 sec)
– Input Power: 35 (50) MW
– Output Power: ~4.5 (500) MW
– Data Rate: 600 GB/day (2PB/day)
The Science
Data Scales in Fusion
1
10
100
1000
10000
100000
1000000
10000000
100000000
MAST JET 2007 JET 2017 ITER DEMO
VolumeofData(GB/Day)
Daily Data Volumes
Difference between LHC/SKA and ITER
Concept Design BuildModel Data
Science
Concept Design BuildModel Data Engine
ering
• Address part of the computing problem
– Scalability
• Cloud Bursting
– Running MPI and Open-MP applications
• In containers
– Makes use of a number of INDIGO components EGI FedCloud
Resources
• IM, Orchestrator, udocker,
– Tested with Geant-4 and Serpent
• Very Successful…
EOSCpilot
EOSCpilot Resource Usage
EOSCpilot Results
Image Courtesy of Andrew Lahiff, CCFE Image Courtesy of Jonathan Shimell, CCFE
• Combining Computing AND Data
– Increasing ITER Relevance
• DATA
– Data Federation and Distribution
• OneData, B2SAFE and Dynafed
• Data Readable and Writable using ITER middleware
• Computing
– Testing Cloud Deployment of IMAS
• Extending work of EOSCpilot
• AAI
– Integrating Proposed Fusion AAI into EOSC
EOSC-hub
• EOSCpilot has allowed us to successfully demonstrate the
usage of cloud computing for HPC work
– Both MPI and Open-MP
• EOSC-Hub will allow us to demonstrate
– ITER relevance of the work done on pilot
– Support EUROFusion Open Data policies
– Integrate Data Federation and Cloud Computing
Summary
• We would like to thank the following for use of cloud
resources in EOSCpilot:
– EGI Federation, CESGA, CESNET, IN2P3, ReCaS-Bari
• And the support provided by INDIGO-Datacloud in making the
demonstrator a success
Acknowledgements

Shaun de Witt: Example for an EOSC-hub Competence Centre example - Fusion

  • 1.
    CCFE is thefusion research arm of the United Kingdom Atomic Energy Authority. This work was part-funded by the RCUK Energy Programme [grant number EP/I501045] and the European Union’s Horizon 2020 research and innovation programme. Small Scale HPC for Big Science Shaun de Witt shaun.de-witt@ukaea.uk EOSC-Hub Public Day Malaga, April 2018
  • 2.
    • Background • EOSCpilotWork and Preliminary Results • EOSC-Hub Planned Work • Conclusions Agenda
  • 3.
    • Making aSun on Earth – Fusing nuclei together – Brings together compute intensive disciplines • UQ, FEM, Radiative Transfer, Neutronics, spin lattice dynamics, • The Numbers – JET/ITER: – Plasma Temperature: >100x106 (150x106) C – Plasma Density: ~10-6 (10-5) atmospheric – Duration of ‘shot’: 5-20 sec (350 sec) – Input Power: 35 (50) MW – Output Power: ~4.5 (500) MW – Data Rate: 600 GB/day (2PB/day) The Science
  • 4.
    Data Scales inFusion 1 10 100 1000 10000 100000 1000000 10000000 100000000 MAST JET 2007 JET 2017 ITER DEMO VolumeofData(GB/Day) Daily Data Volumes
  • 5.
    Difference between LHC/SKAand ITER Concept Design BuildModel Data Science Concept Design BuildModel Data Engine ering
  • 6.
    • Address partof the computing problem – Scalability • Cloud Bursting – Running MPI and Open-MP applications • In containers – Makes use of a number of INDIGO components EGI FedCloud Resources • IM, Orchestrator, udocker, – Tested with Geant-4 and Serpent • Very Successful… EOSCpilot
  • 7.
  • 8.
    EOSCpilot Results Image Courtesyof Andrew Lahiff, CCFE Image Courtesy of Jonathan Shimell, CCFE
  • 9.
    • Combining ComputingAND Data – Increasing ITER Relevance • DATA – Data Federation and Distribution • OneData, B2SAFE and Dynafed • Data Readable and Writable using ITER middleware • Computing – Testing Cloud Deployment of IMAS • Extending work of EOSCpilot • AAI – Integrating Proposed Fusion AAI into EOSC EOSC-hub
  • 10.
    • EOSCpilot hasallowed us to successfully demonstrate the usage of cloud computing for HPC work – Both MPI and Open-MP • EOSC-Hub will allow us to demonstrate – ITER relevance of the work done on pilot – Support EUROFusion Open Data policies – Integrate Data Federation and Cloud Computing Summary
  • 11.
    • We wouldlike to thank the following for use of cloud resources in EOSCpilot: – EGI Federation, CESGA, CESNET, IN2P3, ReCaS-Bari • And the support provided by INDIGO-Datacloud in making the demonstrator a success Acknowledgements

Editor's Notes

  • #4 Atruistically – Make sure data is preserved and usable for future generations of scientists and engineers Legally – we want to be able to prove that our results are correct
  • #10 IMAS – the ITER Integrated Modelling and Analysis Suite