Computational fluid dynamics
Transcription
Computational fluid dynamics
Computing Services at CSC Janne Ignatius Director Computing Services CSC Fact Sheet Operated on a non-profit principle All shares to the Ministry of Education of Finland in 1997 CSC Turnover in 2006 15.6 M€, 170 employees Since March 2005, facilities in Keilaniemi, Espoo Reorganized as a limited company, CSC-Scientific Computing Ltd. in 1993 First supercomputer Cray X-MP/EA 416 in 1989 Founded in 1971 as a technical support unit for Univac 1108 Funet started in 1984 MISSION: CSC, as a part of the Finnish national research structure, develops and offers high quality information technology services VISION 2012: CSC – a leading center of excellence in information technology for science in the European research area CSC supports the national research structure CSC Fields of Services FUNET FUNETSERVICES SERVICES UNIVERSITIES UNIVERSITIES POLYTECHNICS POLYTECHNICS RESEARCH RESEARCHINSTITUTES INSTITUTES COMPANIES COMPANIES COMPUTING COMPUTINGSERVICES SERVICES DATA DATASERVICES SERVICES APPLICATION APPLICATIONSERVICES SERVICES INFORMATION INFORMATIONMANAGEMENT MANAGEMENT SERVICES SERVICES Managing Managingdirector director Director,customer customerservices services Director, Customer Solutions and processes HR and office services Communications Funet Funetnetwork networkservices services Information Informationmanagement managementservices services Data Dataservices servicesfor forscience scienceand andculture culture Application Applicationservices services Computing Computingservices services Finance services General administration Computing Services COMP Janne Ignatius Core Computing Support CORE Computing Environment and Applications CEA Special Computing SPECO Tero Tuononen Jussi Heikonen Jura Tarus [Juha Helin (Cray)] Petri Isopahkala Esko Keränen Kari Kiviaura Jari Niittylahti Aila Lyijynen Samuli Saarinen Dan Still Joni Virtanen Sebastian von Alfthan Tommi Bergman Jussi Enkovaara Pekka Manninen Marko Myllynen Jarmo Pirhonen Raimo Uusvuori Thomas Zwinger Juha Fagerholm Juha Jäykkä Vesa Kolhinen Olli-Pekka Lehto Juha Lento Petri Nikunen Nino Runeberg Customers ● 3000 researchers use CSC’s computing capacity ● Funet connects 85 organisations to the global research networking infrastructure – All Finnish universities – Nearly all Finnish polytechnics – 30 industrial clients and research institutions – Total of 350 000 end users Users by discipline 2007 220 286 Biosciences Computer science 50 Physics 54 Chemistry 59 Linguistics Nanoscience 112 211 Computational fluid dynamics Engineering 121 152 203 Mathematics Other Usage of processor time by discipline 2007 1% 2% 1% 2 %1 % 2% Nanoscience Physics 6% 38 % 8% Chemistry Biosciences Astrophysics Computational fluid dynamics Computer science Earth sciences 14 % Environmental sciences Computational drug design 25 % Other Expert Services in Computational Science • • • • • • • • • • • Biosciences Geosciences Physics Chemistry Nanoscience Linguistics Numerics Parallel Computing Structural analysis Computational fluid dynamics Visualization Performance pyramid European HPC center(s) National/regional centers Local centers TIER 0 TIER 1 TIER 2 The ESFRI Vision for a European HPC service European HPC-facilities at the top of an tier-0 HPC provisioning pyramid – Tier-0: 3-5 European Centres – Tier-1: National Centres – Tier-2: Regional/University Centres tier-1 PRACE DEISA EGEE tier-2 Creation of a European HPC ecosystem involving all stakeholders – – – – HPC service providers on all tiers Grid Infrastructures Scientific and industrial user communities The European HPC hard- and software industry 13 PRACE – Project Facts Objectives of the PRACE Project: – Prepare the contracts to establish the PRACE permanent Research Infrastructure as a single Legal Entity in 2010 including governance, funding, procurement, and usage strategies. – Perform the technical work to prepare operation of the Tier-0 systems in 2009/2010 including deployment and benchmarking of prototypes for Petaflops systems and porting, optimising, peta-scaling of applications Project facts: – Partners: 16 Legal Entities from 14 countries – Project duration: January 2008 – December 2009 – Project budget: 20 M € , EC funding: 10 M € PRACE is funded in part by the EC under the FP7 Capacities programme grant agreement INFSO-RI-211528 14 Site Architecture Point of contact FZJ MPP Germany IBM BlueGene/P Michael Stefan m.stephan@fzjuelich.de CSC-CSCS MPP Finland+Switzerland Cray XT5 (/XTn) - AMD Opteron Janne Ignatius janne.ignatius@csc.fi Peter Kunszt peter.kunszt@cscs.ch CEA-FZJ SMP-TN France+Germany Bull et al. Intel Xeon Nehalem Gilles Wiber gilles.wiber@cea.fr Norbert Eicker n.eicker@fz-juelich.de NCF SMP-FN Netherlands IBM Power 6 Axel Berg axel@sara.nl Peter Michielse michielse@nwo.nl BSC Hybrid – fine grain Sergi Girona sergi.girona@bsc.es Spain IBM Power6 + Cell HLRS Hybrid – coarse grain Stefan Wesner wesner@hlrs.de Germany NEC Vector SX/9 + x86 DEISA –Distributed European Infrastructure for Supercomputing Applications A consortium of leading national supercomputing centres deploying and operating a persistent, production quality, distributed supercomputing environment with continental scope Grid-enabled FP6 funded Research Infrastructure A 4+3-year project started in May 2004 Total budget is 37,1 M€ (incl. DEISA and eDEISA contracts), EU funding - 20.9 M€ EGEE-II Applications Overview Enabling Grids for E-sciencE • • >200 VOs from several scientific domains – Astronomy & Astrophysics – Civil Protection – Computational Chemistry – Comp. Fluid Dynamics – Computer Science/Tools – Condensed Matter Physics – Earth Sciences – Fusion – High Energy Physics – Life Sciences Further applications under evaluation Applications have moved from testing to routine and daily usage ~80-90% efficiency EGEE-II INFSO-RI-031688 98k jobs/day Layers of Computing Capacity for Science Tier 0: European Top-level Computing Capacity (under development) 0 1 2 Tier 1: National level: supercomputer system at CSC i) more tightly connected part (more expensive per CPU) ≈ ’capability’ part ii) massive cluster part (less expensive per CPU) ≈ ’capacity’ part Tier 2: (Cross-)departmental clusters at universities and research institutes Supercomputer ‘Louhi’: Cray • Cray XT4/XT5 – genuine homogeneous massively parallel processor (MPP) system – Cray’s proprietary interconnect SeaStar2(+), very good scalability • At all phases – memory 1 GB/core, but in a subset of cores 2 GB/core (in 384 XT4 and 768 XT5 cores) – local fast work disk 70 TB (fiber channel) • Operating system: – in login nodes complete Linux (Unicos/lc, based on SuSE Linux) – in computational cores Cray Linux Environment: a light-weight Linux kernel which requires certain minor modifications to the source code of programs Supercomputer ‘Louhi’: Cray Peak performance 10.5 Tflops -> 86 Tflops • Phase 1 – In full operation since April 2, 2007 – 11 production cabinets, AMD Opteron 2.6 GHz 2-core, 10.5 Tflops peak performance, 2024 computational cores • Phase 2 – At 2 stages in summer and fall 2008, 2.3 GHz AMD Opteron 4core – Upgrading in XT4 the dual-core CPUs to quad-core CPUs, and bringing in XT5 additional cabinets; single combined system – The combined XT4/XT5 configuration has been in general customer usage since October 8, 2008: 86.7 Tflops peak power, 9424 computational cores Supercluster ‘Murska’: HP Peak performance 11.3 Tflops • • • • • • • In full operation since July 12, 2007 2176 processor cores: 2.6 GHz AMD Opteron 2-core Interconnect: Infiniband (industry standard, middle-range scalability) Local fast work disk: 98 TB (SATA) Memory: in 128 core 8 GB/core (32 GB/node), in 512 core 4 GB/core, in 512 core 2 GB/core, in 1024 core 1 GB/core. HP XC-cluster, operating system is Linux (based on Red Hat Enterprise Linux Advanced Server 4.0) Machine room and hosting provided by HP for at least the first 2 years On profiling the systems • HP supercluster ‘Murska’: – – – – – For a large number of computational customers No porting of codes needed Data-intensive computing Jobs needing more memory ‘Multiserial’ jobs • Cray XT4/XT5 supercomputer ‘Louhi’: – For a smaller number of research groups – For runs/codes which benefit from Cray (in performance/price – sense) – Software: only as needed by those groups – Customers will be helped in porting the codes to Cray – In batch queue system more towards capability-type profiling: space for very large parallel jobs => a somewhat lower usage percentage CSC’s history of supercomputers Image: Juha Fagerholm, CSC CSC / Top500 sum CSC's share of Top500-sum 0.008 0.006 0.004 0.002 0 1995 2000 Year 2005 [(c) CSC, J. Ignatius] 2010 CSC / Top500 sum CSC's share of Top500-sum 0.008 0.006 0.004 0.002 0 1995 2000 Year 2005 [(c) CSC, J. Ignatius] 2010 Layers of Computing Capacity for Science Tier 0: European Top-level Computing Capacity (under development) 0 1 2 Tier 1: National level: supercomputer system at CSC Tier 2: (Cross-)departmental clusters at universities and research institutes Material Sciences National Grid Infrastructure (M-grid) • • • a joint project of CSC, 7 Finnish universities and Helsinki Institute of Physics, funded partially by the Finnish Academy in the National Research Infrastructure Programme has built a homogeneous PC-cluster environment with theoretical peak of approx. 3 Tflop/s (350 nodes) Environment – Hardware: Dual AMD Opteron 1.8-2.2 Ghz nodes with 2-8 GB memory, 1-2 TB shared storage, separate 2xGE (communications and NFS), remote administration – OS: NPACI Rocks Cluster Distribution / 64 bit, based on RedHat Enterprise Linux 3 – Grid middleware: NorduGrid ARC Grid MW compiled – With Globus 3.2.1 libraries, Sun Grid Engine as LRMS