Your SlideShare is downloading. ×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Saving this for later?

Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime - even offline.

Text the download link to your phone

Standard text messaging rates apply

dr. ross nobes - Innovation in High-Performance Computing

1,022
views

Published on

Published in: Business, Technology

0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
1,022
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
28
Comments
0
Likes
1
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1.
  • 2. Innovation Forum – Room 12
    Innovation in High-Performance Computing
    Dr. Ross Nobes
    Assistant Division Manager, Environment and Health
    Fujitsu Laboratories
    9:00 h
  • 3. TRAILER 1
    1m27s
  • 4. Innovation in High-Performance Computing
    Ross Nobes
    Fujitsu Laboratories of Europe
    4
    Copyright 2010 FUJITSU
  • 5. What is High Performance Computing?
    5
    Use of parallel computers with low-latency, high-bandwidth networks
    Capability computing
    Large, complex simulations
    Climate modelling, advanced materials, renewable energy, nuclear fusion, aerospace industry, national security, basic science, etc.
    Data-intensive computing
    Large HadronCollider, Square Kilometre Array
    Now in “petascale” era (1015 operations per second)
    Moving towards “exascale” computing (1018) by ~2020
    Capacity computing
    Large numbers of smaller compute-intensive jobs
    Digital content creation in entertainment industry, financial applications, etc.
    Copyright 2010 FUJITSU
  • 6. Innovation in Hardware
    The “K Computer”
    6
  • 7. 30+ Years of Supercomputing
    7
    Copyright 2010 FUJITSU
  • 8. The Next-Generation Supercomputer
    8
    RIKEN and Fujitsu are jointly developing Japan’s Next-Generation Supercomputer, the “K Computer”
    10 PFLOP/s performance
    Installed at RIKEN’s Advanced Institute for Computational Science, Kobe
    The centrepiece of Japan’s new “Innovative High-Performance Computing Infrastructure”
    Copyright 2010 FUJITSU
    Tuning and improvement
    Production, installation
    and adjustment
    Prototype, evaluation
    Detailed design
    Conceptual design
  • 9. The K Computer
    9
    Copyright 2010 FUJITSU
    Fujitsu Begins Shipping Japan's Next-Generation Supercomputer
    Tokyo, September 28, 2010 — Fujitsu announced that today it began shipping the computing units for Japan's Next-Generation Supercomputer, nicknamed the "K computer"
  • 10. New Technologies
    10
    Copyright 2010 FUJITSU
    • Tofu Interconnect
    System
    :
    • VISIMPACT Technology
    • 11. Advanced Packaging
    Node
    CPU
    :
    Core
    Core
    • SPARC64TM VIIIfx Multi-Core CPU
    • 12. HPC-ACE Instruction Set
    • 13. Low Power Consumption
    Core
  • 14. Fujitsu’s Advanced Technology
    11
    New HPC-enhanced SPARC64TM VIIIfx CPU
    8 cores, 2 GHz, 128 GFLOP/s
    58 W peak, 2.2 GFLOP/s per watt
    Water cooling
    Low current leakage
    Low power consumption and low failure rate
    Highlyreliabledesign
    Specific logic circuits to detect and correct errors
    VISIMPACT
    Combines hardware enhancements and advanced compiler technologies to promote hybrid programming models
    Copyright 2010 FUJITSU
  • 15. VISIMPACT: Performance
    12
    Copyright 2010 FUJITSU
    LS-DYNA on Fujitsu FX1 Supercomputer
    www.topcrunch.org
  • 16. Effect of VISIMPACT
    13
    Copyright 2010 FUJITSU
    • Test Case: Caravan 10-million Elements Car-to-Car Model
    • 17. Time Duration: 10 milliseconds ->120 milliseconds
    1.54 times speedup by VISIMPACT
  • 18. High-Performance Computing
    Opportunities for Europe
    14
  • 19. IDC Report
    15
    HPC use is indispensable for advancing both scientific and industrial competitiveness
    Europe is under-investing in HPC, while other nations are growing their supercomputer investments dramatically
    The transition to petascale and exascale computing creates opportunities: For Europe to return to the forefront of development for the next generation of research and HPC software/hardware/storage/networking technologies
    Copyright 2010 FUJITSU
  • 20. IDC Vision for Europe
    16
    Copyright 2010 FUJITSU
    Provide World-Class HPC Expertise and Resources
    To Make EU Scientists, Engineers And Analysts
    The Most Productive and Innovative In The World
    In Applying HPC To Advance Their Research
    In The Pursuit Of Scientific Advancement And Economic Growth
  • 21. IDC: Outcomes from the Vision
    17
    Copyright 2010 FUJITSU
    By the year 2020, the EU HPC strategy has enabled the following progress to occur:
    Europe is recognised as a hotbed for new science and engineering research
    Europe's leadership in the targeted areas has created many new jobs in science and industry, and has caused the national economies to grow faster
    Europe is the world leader in important high-end HPC technologies
    Europe leads the world in scalable algorithms and software applications, and in tools to make HPC systems easy-to-use and to make researchers highly innovative and productive
    FLE’s target
  • 22. Innovation in Software
    The Open Petascale Libraries Project
    18
  • 23. Open Petascale Libraries Project
    19
    Copyright 2010 FUJITSU
    Global collaboration to develop advanced numerical software for supercomputing
    Initiated by Fujitsu Laboratories of Europe and Fujitsu Limited
    Dedicated forum to promote the open exchange of ideas and the collaborative development of general-purpose and application-specific numerical libraries
    Targeted initially at parallel computers built from multi-core processors
    All output available as open-source software
    Official launch of OPL in November 2010 to coincide with SC10 conference in New Orleans
  • 24. Exascale Application and Data Initiative
    20
    Copyright 2010 FUJITSU
    Computational science society
    Open software communities
    Exascale
    Application
    and
    Data
    Initiative
    FUJITSU
    Platform Solution
    R&D Labs
    Application users
    Application developers
    Computational scientists,
    computer scientists and engineers
    Vendors
  • 25. Bridging Applications and Platforms
    21
    Copyright 2010 FUJITSU
  • 26. OPL: Structure
    22
    Copyright 2010 FUJITSU
    Advisory Panel
    Eminent figures in HPCand numerical algorithms
    EESI
    Steering Committee
    formed from representative of Network members
    Japanese Members
    Global Members
    Secretariat:
    Fujitsu Labs Europe / Fujitsu Ltd
    Open Petascale Libraries Network
  • 27. OPL: Members
    23
    Copyright 2010 FUJITSU
  • 28. Contributing to Many Application Areas
    24
    Copyright 2010 FUJITSU
  • 29. www.openpetascale.org
    For More Information
    25
    Copyright 2010 FUJITSU
  • 30. Innovation in Software
    Case Study: Improving the Pharmaceutical R&D Pipeline
    26
  • 31. Modelling in the Pharmaceutical Industry
    27
    We no longer design aeroplanes, bridges, cars or large buildings without first simulating them using computer models
    Drugs are developed without modelling their interactions with our bodies
    This results in a very expensive pharmaceutical research process
    Many failures at late stages of the testing process
    High health care costs
    Copyright 2010 FUJITSU
  • 32. Lifecycle of a New Medicine
    Patent application
    Acute toxicity
    Phase I clinical trials
    Pharmacology
    Chronic toxicity
    Registration and transparency
    Phase II
    Phase III
    Pharmacovigilance
    Reimbursement
    Price
    1 medicinal product
    0
    5 years
    10 years
    15 years
    20 years
    Patent expiry
    SPC
    (supplementary
    protection certificate)
    max. + 5 years
    10 years of research
    2 to 3 years of administrative
    procedures
    • Average R&D cost for a new medicine: €1 billion
    • 33. Only 3 out of 10 drugs cover R&D costs before patent expiry
    28
    Copyright 2010 FUJITSU
  • 34. Pharmaceutical R&D
    29
    From lab to shelf, development of a new medicine takes 10-13 years
    For every 5000 molecules tested:
    250 enter pre-clinical testing
    10 enter clinical testing
    1 approved for use in patients
    Half of drugs that enter Stage III clinical testing fail at that stage
    Often from side-effects in the heart (arrhythmia and fibrillation which can lead to heart failure and death)
    The European project preDiCT aims to eliminate unsuitable drugs early in the R&D pipeline to eliminate wasted time and expenditure
    Copyright 2010 FUJITSU
  • 35. preDiCT Partners
    30
    Copyright 2010 FUJITSU
  • 36. Objectives of preDiCT
    31
    Earlier approval of successful drugs, at lower cost
    Early elimination of unsuitable drug candidates
    Systematic inclusion of modelling as well as experiments in the drug testing process
    Real time simulation of cardiac electrical activation on a realistic whole human heart mesh
    Modelling environment suitable for use by physiologists
    Better cardiac cell models to describe the interaction of healthy and unhealthy tissue with drugs
    Identification of simple biomarkers to indicate the risk of a drug inducing arrhythmia
    Copyright 2010 FUJITSU
  • 37. Whole-Ventricle Simulator
    Chaste
    Optimized cell
    code
    Scale 2
    Cell modelling
    CellML
    Requirements
    Scale 3
    Organ modelling
    Fast, parallel
    simulator
    Research tool
    32
    Copyright 2010 FUJITSU
  • 38. Technological Improvements
    33
    Copyright 2010 FUJITSU
  • 39. Performance Improvements
    34
    Copyright 2010 FUJITSU
  • 40. Virtual Physiological Human
    35
    preDiCT is part of the VPH Initiative, a flagship eHealth project in FP7
    Computational frameworks and ICT-based tools for multiscale models of the human anatomy, physiology and pathology
    A key facilitator for:
    Personalised (patient-specific) healthcare solutions
    Early diagnosis and predictive medicine
    Assessment of safety/efficacy of drugs using patient-specific computational models
    Innovative personalised drugs
    FLE plans to contribute further to the VPH
    Copyright 2010 FUJITSU
  • 41. Future Innovation
    The International Exascale Software Project
    36
  • 42. Assumed Exascale Architectures
    37
    Copyright 2010 FUJITSU
  • 43. Assumed Exascale Architectures
    38
    Copyright 2010 FUJITSU
    x1000
    x3
    x100
    x10
    x10
    O(1000)
    O(1,000,000)
    x100
    x0.1
  • 44. Challenges for Exascale
    39
    Copyright 2010 FUJITSU
    Will need 1 billion simultaneous threads of execution
    -> Daunting challenge for algorithm scalability
    Power budget will be limited to 20 MW
    -> Need for energy efficient algorithms
    A node will fail every few hours
    -> Need for fault tolerant algorithms
    Fujitsu Laboratories intends to work with the computational science community to tackle these challenges
    International Exascale Software Project
  • 45. Summary
    40
    High-performance computing is a vital tool for future scientific and industrial competitiveness and economic prosperity
    Fujitsu Limited is developing innovative HPC systems, such as the K Computer
    Innovation in software is a challenge that Fujitsu Laboratories is tackling through open collaboration
    Numerical libraries
    New applications, especially in health, energy and the environment
    The road ahead to exascale computing
    Copyright 2010 FUJITSU