Published on September 25, 2014
Japan’s Policy toward Exascale Computing Yoshio KAWAGUCHI Office for Promotion of Computing Science / MEXT 27 February, 2014
1. Current HPC Status in Japan 2. Japanese Policy for Development of HPC Systems 1 Today’s Topics
Current HPC Status in Japan 2
High Performance Computing Infrastructure (HPCI) Tohoku U. U. Tsukuba Hokkaido U. HPCI is a nation-wide HPC infrastructure into which the supercomputers and large scale storages are mutually connected through the high speed network, SINET4. Kyoto U. Osaka U. “K” computer Kyushu U. JAMSTEC Nagoya U. Riken AICS • “K” computer • Shared Storage • Supercomputers • Shared Storage NII • Management of SINET &Single sign-on Tokyo Tech U. Tokyo 3
The building of “K” and AICS “K computer (on July. 2012) General Use Category ~30% （To be called for proposals） General Use projects ~20% Industrial Use Projects ~ 5% Junior Researcher Promotion Projects ~ 5% Computing Resources Allocated for System Enhancement ~15% Strategic Program Category ~50% Additional Allocation for acceleration of achievements ~5% Number of K users 1,431 Number of projects on K Computer 136 Rack 4 General Use Category of the K Computer
Life science/Drug design New materials/New energy MONODUKURI (Manufacturing technologies) Global climate change prediction for disaster prevention/mitigation Toshio YANAGIDA RIKEN Shinji TSUNEYUKI University of Tokyo Shiro IMAWAKI JAMSTEC Chisachi KATO University of Tokyo Shinya AOKI University of Tsukuba creation The origin of matters and the universe Strategic organizaions 5 Strategic Application Areas
0%# 10%# 20%# 30%# 40%# 50%# 60%# 70%# 80%# 90%# 100%# 9/28.10/7# 10/8.10/14# 10/15.10/21# 10/22.10/28# 10/29.11/2# 11/8.11/11# 11/12.11/18# 11/19.11/25# 11/26.12/2# 12/3.12/9# 12/10.12/16# 12/17.12/23# 12/24.12/30# 12/31.1/6# 1/7.1/13# 1/14.1/20# 1/21.1/27# 1/28.2/3# 2/4.2/10# 2/11.2/14# 2/19.2/24# 2/25.3/3# 3/4.3/10# 3/11.3/17# 3/18.3/24# 3/25.3/31# 4/1.4/7# 4/8.4/14# 4/15.4/21# 4/22.4/28# 4/29.5/5# 5/6.5/12# 5/13.5/19# 5/20.5/26# 5/27.6/2# 6/3.6/9# 6/10.6/16# 6/17.6/23# 6/24.6/30# 7/1.7/7# 7/8.7/14# 7/15.7/21# 7/22.7/29# 8/5.8/11# 8/12.8/18# 8/19.8/25# 8/26.9/1# 9/2.9/8# 9/9.9/15# 9/16.9/22# 9/23.9/29# 9/30.10/6# 10/7.10/13# 10/14.10/20# 10/21.10/27# 10/28.11/3# 11/4.11/10# 11/11.11/17# 11/18.11/24# 11/25.12/1# 12/2.12/8# 12/9.12/15# 12/16.12/22# 12/23.12/28# 1/2.1/5# 1/6.1/12# 1/13.1/19# 1/20.1/26# 1/27.2/2# 2/3.2/9# 2/10.2/16# 2/17.2/23#
&' #%$% >80000# 50001.80000# 20001.50000# 10001.20000# 5001.10000# 2001.5000# 1001.2000# 501.1000# 101.500# <100# !" 2012 Oct Nov Dec 2013 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2014 Jan Feb 6 non-trouble operation 93.0% scheduled maintenanc e 5.4% irregular service stop 1.6% About 80% of CPU resources are used for jobs The rate of operation is about 93% CPU resources which are occupied to run applications Operation of K Computer
Real space DFT calculations of Silicon nanowires (Gordon Bell Prize 2011) The world’s first nano-level high precision simulation with the actual material size. Applications by Industry for Using K Computer Docking simulation between protein and compound Estimating about 19 billion docking pairs through high speed calculations by the organized 11 Japanese pharmaceutical companies. The gravitational trillion-body problem (Gordon Bell Prize 2012) The world’s largest number of dark matter simulation performed on the K computer. Nanometer level simulation of rubber for tire Contribution to the development of the new rubber material for tires which realizes low fuel consumption and high grip performance. 7 The researchers who conducted simulation researches in K received Gordon Bell Prizes in two consecutive years (2011&2012).
Japanese Policy for Development of HPC Systems 8
○ Comprehensive Strategy on Science, Technology and Innovation (Cabinet Decision) • Promotion of development, establishment, and wide utilization of world-class R&D infrastructure (e.g., utilization of supercomputers); establishment of an environment where skilled talents from the industry, academia, and government can attempt to conduct innovative groundbreaking research projects beyond the field or organizational boundaries. 9 Comprehensive Strategy on Science,Technology and Inovation (June,2013）
Continuous Development of HPCI (High Performance Computing Infrastructure) Peak Performance [FLOPS] FLOPS: FLoating point Operations Per Second 1Exa 100P 10P 1P 100T Continuous Development of Top-class Supercomputers Industry, Laboratories Industry, Laboratories Universities, National Institutes Industry, Laboratories 2000 2010 2015 2020 2025 CP-PACS NWT Universities, National Earth Simulator Institutes Exascale Supercomputer Raise the Level of K-Computer Low-end Supercomputers 10 Universities, National Institutes
＜Picture of Infrastructure for Computational Science and Technology in Japan＞ Flagship System System with performance at world’s top level for wide-area applications Characterized systems supporting Flagship system Aim to develop Exascale-supercomputer Leading Machines by 2020 Systems at Supercomputer Centers in 9 Universities Systems at National Institutes Other Systems in Universities, and so on. Nation-wide Infrastructure Operated by HPCI ©RIKEN 11 Direction for Development of Next Gen. Supercomputers
" Development of new medicine with little side effect Computing simulation that could take the possibility of side effects into account " Prevention and mitigation against wide-ranging complex disaster Damage prediction and mitigation of the wide-ranging complex disaster including earthquake, tsunami, refuge, people’s movement and so on. Propagation " Next-generation device science Predicting the unique physical properties of a composite material in one million atomic levels. " Development of higher safety car Evaluating damages of not only car itself but human body (injury of bones, entrails, etc) of driver. * Black shows injured body parts. " Development of an epoch-making battery The combination of the optimal electrode material and an electrolyte is realized and it contributes to the development of the advanced battery which has high safety and tolerance. " Explorer of the origin and evolution Clarifying the large variety of galaxies and planets by integrated simulations of the overall universe ranging from planets formation to large scale structure formation in the universe. Large scale structure formation galaxy formation star formation * planet formation of the universe ※ Matthew Baw (University of Exeter) EarthquakeRefugePeople's movementRestoration & Reconstruction Tsunami 12 Major Social and Science Challenges to Be Solved Using Exascale Supercomputer
n Double-digits (higher)performance by 2020 n Push state of the art in power efficiency, scalability & reliability n Enable unprecedented application capability n AICS RIKEN in charge of exascale systems development n Total project cost ca. JPY140 billion with about JPY 110 billion from the government’s budget (JPY 1.2 billion for 2014) Year of 2013 2014 2015 l Architecture: general-purpose plus accelerator components l Target performance of 1 Exaflops, 100 times as powerful as K l Power consumption of 30-40MW (cf. K computer: 12.7 MW) Basic Design Manufacture Operation Development and utilization of Application for Exascale Computing System Application Trial Production Detailed Design Installation Coordination Basic design for GP & Accelerator components and thorough review the system specification 2016 2017 2018 2019 2020 Outline: 13 Japan Exascale System Development Schedule:
Project Arrangement between US and Japan on R&D Collaboration for HPC System Software Development The Implementing Arrangement Concerning Cooperation in R&D in Energy and Related Fields MEXT, Japan ó DOE, US <April 30, 2013> ※At Joint High Level Committee Meeting on Science and Technology cooperation (April 30, 2013) ※Cooperative area described in this arrangement: (Nuclear Fusion Science, High Energy Physics, Nuclear Physics, Computer Science, etc) (As One of cooperative area) Project Arrangement to the Implementing Arrangement between US and Japan Concerning Computer Science and Software Related to HPC for Open Scientific Research • Participating Organizations: National Lab. (DOE, US), RIKEN (Japan) • Cooperative Area: System Software • Major Contents Description for Forms of Cooperation and Management of Project etc TBC 14 Agreement between US and Japan on Cooperation in R&D in Sci. and Tech. (1988)
○HPCI（High Performance Computing Infrastructure) ・Started in September 2012 ・K Computer, other supercomputers, large storage, network ・Great success ○Development of Exascale Supercomputer ・Starting point of long and hard way 15 Summary MEXT continuously promotes both projects
Presentación que realice en el Evento Nacional de Gobierno Abierto, realizado los ...
In this presentation we will describe our experience developing with a highly dyna...
Presentation to the LITA Forum 7th November 2014 Albuquerque, NM
Un recorrido por los cambios que nos generará el wearabletech en el futuro
Um paralelo entre as novidades & mercado em Wearable Computing e Tecnologias Assis...
Japan’s Policy toward Exascale Computing ... Development and utilization of Application for Exascale Computing System ... Japan Exascale System ...
View 88 Exascale Computing posts, ... ExaScale Computing Requires ExaScale Thinking. ... Japan's policy toward exascale computing (2014.2)
... and Japan have each moved aggressively to develop their own plans for achieving exascale computing in the next decade.
Exascale computing refers to computing systems ... In Japan, the RIKEN Advanced Institute for Computational Science is planning an exascale system ...
Contemporary High Performance Computing: From Petascale toward Exascale ... Performance Computing: From Petascale toward Exascale ... in Japan, Satoshi ...
Petascale computing is being used to do advanced computations in fields such as weather and climate ... such as Germany and Japan, ... Exascale computing;