2. Top500 List of Supercomputers
H. Meuer, H. Simon, E. Strohmaier, & J. Dongarra
- Listing of the 500 most powerful
Computers in the World
- Yardstick: Rmax from LINPACK MPP
Ax=b, dense problem TPP performance
Rate
- Updated twice a year Size
SC‘xy in the States in November
Meeting in Germany in June
2 - All data available from www.top500.org
4. November 2011: The TOP10
Rmax % of Power MFlops
Rank Site Computer Country Cores
[Pflops] Peak [MW] /Watt
RIKEN Advanced Inst K computer Fujitsu SPARC64
1 Japan 705,024 10.5 93 12.7 826
for Comp Sci VIIIfx + custom
Nat. SuperComputer Tianhe-1A, NUDT
2 China 186,368 2.57 55 4.04 636
Center in Tianjin Intel + Nvidia GPU + custom
DOE / OS Jaguar, Cray
3 USA 224,162 1.76 75 7.0 251
Oak Ridge Nat Lab AMD + custom
Nat. Supercomputer Nebulea, Dawning
4 China 120,640 1.27 43 2.58 493
Center in Shenzhen Intel + Nvidia GPU + IB
GSIC Center, Tokyo Tusbame 2.0, HP
5 Japan 73,278 1.19 52 1.40 850
Institute of Technology Intel + Nvidia GPU + IB
DOE / NNSA Cielo, Cray
6 USA 142,272 1.11 81 3.98 279
LANL & SNL AMD + custom
NASA Ames Research Plelades SGI Altix ICE
7 USA 111,104 1.09 83 4.10 265
Center/NAS 8200EX/8400EX + IB
DOE / OS
Hopper, Cray
8 Lawrence Berkeley Nat USA 153,408 1.054 82 2.91 362
AMD + custom
Lab
Commissariat a
Tera-10, Bull
9 l'Energie Atomique France 138,368 1.050 84 4.59 229
Intel + IB
(CEA)
DOE / NNSA Roadrunner, IBM
10 USA 122,400 1.04 76 2.35 446
Los Alamos Nat Lab AMD + Cell GPU + IB
5. November 2011: The TOP10
Rmax % of Power MFlops
Rank Site Computer Country Cores
[Pflops] Peak [MW] /Watt
RIKEN Advanced Inst K computer Fujitsu SPARC64
1 Japan 705,024 10.5 93 12.7 830
for Comp Sci VIIIfx + custom
Nat. SuperComputer Tianhe-1A, NUDT
2 China 186,368 2.57 55 4.04 636
Center in Tianjin Intel + Nvidia GPU + custom
DOE / OS Jaguar, Cray
3 USA 224,162 1.76 75 7.0 251
Oak Ridge Nat Lab AMD + custom
Nat. Supercomputer Nebulea, Dawning
4 China 120,640 1.27 43 2.58 493
Center in Shenzhen Intel + Nvidia GPU + IB
GSIC Center, Tokyo Tusbame 2.0, HP
5 Japan 73,278 1.19 52 1.40 865
Institute of Technology Intel + Nvidia GPU + IB
DOE / NNSA Cielo, Cray
6 USA 142,272 1.11 81 3.98 279
LANL & SNL AMD + custom
NASA Ames Research Plelades SGI Altix ICE
7 USA 111,104 1.09 83 4.10 265
Center/NAS 8200EX/8400EX + IB
DOE / OS
Hopper, Cray
8 Lawrence Berkeley Nat USA 153,408 1.054 82 2.91 362
AMD + custom
Lab
Commissariat a
Tera-10, Bull
9 l'Energie Atomique France 138,368 1.050 84 4.59 229
Intel + IB
(CEA)
DOE / NNSA Roadrunner, IBM
10 USA 122,400 1.04 76 2.35 446
Los Alamos Nat Lab AMD + Cell GPU + IB
500 IT Service IBM Cluster, Intel + GigE USA 7,236 .051 53
6. Geographical regions
Count Share % Rmax Rpeak Cores
North America 272 54.40% 32923947 48374869 4659645
Eastern Asia 109 21.80% 25868736 38046465 2520930
Western Europe 49 9.80% 8020850 10532996 1173728
Northern Europe 36 7.20% 3652751 5071283 428832
Eastern Europe 11 2.20% 1482188 2519402 126856
Southern Europe 7 1.40% 665279 1047276 60904
Western Asia 6 1.20% 530526.6 808867.6 115540
Australia and New
4 0.80% 353753.5 479797.9 35424
Zealand
South America 2 0.40% 269730 330444.8 37184
South-central Asia 2 0.40% 187910 242995.2 18128
Southern Africa 1 0.20% 61330 74257.9 6336
South-eastern Asia 1 0.20% 52633 98995 9304
Sums
500 100% 74069633.68 107627649.54 9192811
6
7. South America HPC
Rmax Rpeak Power
Rank Site System Cores
(TFlop/s) (TFlop/s) (Kw)
INPE (National Institute for Space Tup - Cray XT6 12-core
49 Research) 2.1 GHz 30720 205.1 258
Brazil Cray Inc.
Galileu - Sun Blade
NACAD/COPPE/UFRJ x6048, Xeon X5560 2.8
Ghz, Infiniband QDR
290 6464 64.6 72.4 430
Brazil Sun Microsystems
7
8. Japanese K Computer
New Linpack run with 705,024 cores at 10.51 Pflop/s (88,128 CPUs) 8
9. China
• First Chinese Supercomputer to
use a Chinese Processor
Sunway BlueLight MPP
ShenWei SW1600 processor, 16 core,
65 nm, fabbed in China
125 Gflop/s peak
In the Top20 with 139,364 cores &
1.07 Pflop/s Peak
• Coming soon, Loongson (Godson)
processor
8-core, 65nm Loongson 3B processor
runs at 1.05 GHz, with a peak
performance of 128 Gflop/s
9
11. Future Computer Systems
♦ Most likely be a hybrid design
Standard multicore chips and accelerator
(GPUs)
♦ Today accelerators are attached
♦ Next generation more integrated
♦ Intel’s MIC architecture “Knights Corner”
48 x86 cores
♦ AMD’s Fusion
Multicore with embedded graphics ATI
♦ Nvidia’s Project Denver plans to develop
an integrated chip using ARM
architecture
11
13. Major Changes to Software &
Algorithms
• Must rethink the design of our
algorithms and software
Another disruptive technology
• Similar to what happened with cluster
computing and message passing
Rethink and rewrite the applications,
algorithms, and software
Data movement is expense
Flop/s are cheap, so are provisioned in
excess
13
14. Critical Issues at Peta & Exascale for
Algorithm and Software Design
• Synchronization-reducing algorithms
Break Fork-Join model
• Communication-reducing algorithms
Use methods which have lower bound on communication
• Autotuning
Today’s machines are too complicated, build “smarts” into
software to adapt to the hardware
• Fault resilient algorithms
Implement algorithms that can recover from failures/bit flips
• Reproducibility of results
Today can’t guarantee this.
15. International Exascale Software Project
Attendees from universities, Steering Committee
research institutes, government, Jack Dongarra, U of Tennessee/Oak
funding agencies, research Ridge National Lab, US
councils, hardware and software Pete Beckman, Argonne Nat. Lab, US
vendors, industry Franck Cappello, INRIA, FR
Thom Dunning, NCSA, US
Thomas Lippert, Jülich Supercomputing
Centre, DE
Satoshi Matsuoka, Tokyo Inst. of Tech, JP
Paul Messina, Argonne Nat. Lab, US
Patrick Aerts, Netherlands Organization
for Scientific Research, NL
Anne Trefethen, Oxford, UK
Mateo Valero, Barcelona
Supercomptuing Ceneter, Spain
16. International Exascale Software Project Objectives
To enable the international HPC community to improve,
coordinate and leverage their collective investments and
development efforts.
To develop a plan for producing a software infrastructure
capable of supporting exascale applications
Thorough assessment of needs, issues and strategies
Develop a coordinated software roadmap
Provide a framework for organizing the software research
community
Engage vendors to coordinate on how to deal with anticipated
scale
Encourage and facilitate collaboration in education and training
16
17. What Next?
Moving from “What to Build” to “How to Build”
Technology
Defining and developing the roadmap for software
and algorithms on extreme-scale systems
Assessing the short-term, medium-term and long-term
software and algorithm needs of applications for
peta/exascale systems
www.exascale.org
18. What Next?
Moving from “What to Build” to “How to Build”
Organization
Exploring ways for funding agencies to coordinate their
support so that they complement each other
Exploring how laboratories, universities, and vendors
can work together on coordinated HPC software
Creating a plan for working closely with HW vendors
and application teams to co-design future architectures
www.exascale.org
19. What Next?
Moving from “What to Build” to “How to Build”
Execution
Developing a strategic plan for moving forward
Creating a realistic timeline for constructing key
organizational structures and achieving initial goals
Exploring community development techniques and risk
plans to ensure key components are delivered on time
www.exascale.org
20. US TeraGrid
An instrument that delivers high-end IT
resources/services
a computational facility – over two PFlops
Science Gateways –discipline-specific web-portal front-ends
a data storage and management facility – 20 PetaBytes
a high-bandwidth national data network
Support, education and training events
Available freely to research and education projects
with a US lead
21. TeraGrid Objectives
DEEP Science: enabling terascale and petascale science
make science more productive through an integrated set of very-
high capability resources
address key challenges prioritized by users
WIDE Impact: empowering communities
bring TeraGrid capabilities to the broad science community
partner with science community leaders
OPEN Infrastructure, OPEN Partnership
a coordinated, general purpose, reliable set of services and
resources
partner with campuses and facilities
22. The eXtreme Digital (XD) Program
XD : third generation TeraGrid program
2002-2005: Distributed/Extended Terascale Facility
2005-2011: Grid Infrastructure + Resource Providers
2010-2016: eXtreme Digital (XD) + Service Providers
23. 11 Resource Providers, One Facility
UW Grid Infrastructure Group
(UChicago)
UC/ANL PSC
NCAR PU
NCSA
Caltech IU UNC/RENCI
ORNL
USC/ISI NICS
SDSC
LONI
TACC
Resource Provider (RP)
Software Integration Partner
Network Hub
24. eXtreme Digital Resources
High-Performance Computing and Storage Services
High-Performance Remote Visualization and Data Analysis
Services
2 awards; 5 years; $3M/year
Integrating Services (5 years, $26M/year)
Coordination and Management Service (CMS)
5 years; $12M/year
Technology Audit and Insertion Service (TAIS)
5 years; $3M/year
Advanced User Support Service (AUSS)
5 years; $8M/year
Training, Education and Outreach Service (TEOS)
5 years, $3M/year
25. XSEDE : Governance
Leadership
led by NCSA, NICS, PSC, TACC and SDSC: centers with deep
experience
partners who strongly complement these centers with expertise in
science, engineering, technology and education
Balanced governance model
strong central management, rapid response to issues and
opportunities
delegation and decentralization of decision-making authority
openness to genuine stakeholder participation
stakeholder engagement, advisory committees
improved professional project management practices
formal risk management and change control
25
26. XSEDE: Extending Impact
Coordinated national program with greater scope and scale
increased diversity of topics, modes of delivery, and reach to new
communities and audiences
broaden participation among under-represented communities
Campus bridging for effective use of resources
more tightly integrate with campuses through expanded
Champions program and additional bridging activities
Establish certificate and degree programs
institutional incorporation of CS&E curricula; professional
development certificate
prepare undergraduates, graduates and future K-12 teachers
26
27. Summary HPC
Increasingly indispensably to scientific progress and
economy competitiveness
Industrial competiveness ->time to market
National security
Quality of human life
Key element for the competiveness of knowledge
based economies
Not HPC Leadership but innovation leadership
www.exascale.org