Your SlideShare is downloading. ×
DEISA Distributed European Infrastructure for Supercomputing Applications
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Introducing the official SlideShare app

Stunning, full-screen experience for iPhone and Android

Text the download link to your phone

Standard text messaging rates apply

DEISA Distributed European Infrastructure for Supercomputing Applications

810
views

Published on

Published in: Technology

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
810
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
21
Comments
0
Likes
0
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. DEISA Distributed European Infrastructure for Supercomputing Applications Wolfgang Gentzsch gentzsch@rzg.mpg.de ANNUAL HIGH PERFORMANCE COMPUTING www.deisa.eu AND COMMUNICATIONS CONFERENCE Newport RI, March 24-26 2009 RI-222919
  • 2. Partners DEISA: May 1st, 2004 – April 30th, 2008 Three new partners joined June 2005 (eDEISA) DEISA2: May 1st, 2008 – April 30th, 2011 HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 2 RI-222919
  • 3. Partners BSC Barcelona Supercomputing Centre Spain CINECA Consortio Interuniversitario per il Calcolo Automatico Italy CSC Finnish Information Technology Centre for Science Finland EPCC University of Edinburgh and CCLRC UK ECMWF European Centre for Medium-Range Weather Forecast UK (int) FZJ Research Centre Juelich Germany HLRS High Performance Computing Centre Stuttgart Germany IDRIS Institut du Développement et des Ressources France en Informatique Scientifique - CNRS LRZ Leibniz Rechenzentrum Munich Germany RZG Rechenzentrum Garching of the Max Planck Society Germany SARA Dutch National High Performance Computing Netherlands KTH Kungliga Tekniska Högskolan Sweden CSCS Swiss National Supercomputing Centre Switzerland JSCC Joint Supercomputer Center of the Russian Russia Academy of Sciences HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 3 RI-222919
  • 4. The DEISA Supercomputing Grid AIX distributed super-cluster 10 G b Vector systems G E (NEC, …) A N T2 /N R E N Linux systems s (SGI, IBM, …) HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 4 RI-222919
  • 5. Objectives of the European DEISA1 Project • Enabling Europe’s terascale science by integrating Europe’s most powerful HPC systems DEISA is a European Supercomputing Service built on top of existing national HPC services This Service is based on the deployment and operation of a persistent, production quality, distributed supercomputing environment with continental scope • Only Criterion for Success: Enabling scientific discovery across a broad spectrum of science and technology The integration of national facilities and services, together with innovative operational models, is expected to add substantial value to existing infrastructures HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 5 RI-222919
  • 6. Objectives of the European DEISA2 Project • Enhancing the existing distributed European HPC (DEISA1) environment towards a turnkey operational infrastructure • Advancing computational sciences in Europe by supporting user communities and extreme computing projects • Enhancing service provision by offering a manageable variety of options of interaction with computational resources • Integration of T1 and T0 centers • The Petascale T0 Systems need a transparent access from and into the national data repositories • Bridging worldwide HPC and Grid projects HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 6 RI-222919
  • 7. How is DEISA Enhancing HPC Services in Europe? • Running large parallel applications in individual sites, orchestrating the global workload, or by job migration services • Enabling workflow applications with UNICORE (complex applications that are pipelined over several computing platforms), enabling coupled multi-physics Grid applications • Providing a global data management service whose main objectives are: - Integrating distributed data with distributed computing platforms - Enabling efficient, high performance access to remote datasets with Global File System and striped GridFTP. - Integrating hierarchical storage mgmt and databases in the Grid. • Deploying portals to hide complex environments to new users, and to interoperate with other existing grid infrastructures. HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 7 RI-222919
  • 8. HPC Users Requirements • Remote, Standard, Easy Access HPC users: conservative, no interest in complicated middleware stack • Global Login Personal Login in each system, I.e. single “European” username, uniform access • Comfortable Data Access Global, fast and comfortable access to their data, across the centres • Common Production Environment No need for an identical but for an equivalent HPC software stack • Global Help Desk One central point of contact and as local as possible • Application Support Need help in scalability, performance and adaptation to different architectures Sustainability? Carefully listen to the top user value propositions and go build and ‘sell‘ it! HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 8 RI-222919
  • 9. DEISA Services Multiple Common Workflow Presentation ways to production managem. layer access environm. Co- Single Job DEISA reservation Job manag. monitor layer and system a service-and co- rerouting oriented monitor. allocation production (RAS) e-Infrastructure Data WAN Data staging Data manag. transfer shared tools layer tools filesystem Network Unified DEISA Network and AAA Sites connectivity AAA HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA layers9 RI-222919
  • 10. DEISA Core Infrastructure and Services Dedicated high speed network Common AAA Single sign on Accounting/budgeting Global data management High performance remote I/O and data sharing with global file systems High performance transfers of large data sets User Operational infrastructure Distributed Common Production Environment (DCPE) Job management service Common user support and help desk System Operational infrastructure Common monitoring and information systems Common system operation Global Application Support HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 10 RI-222919
  • 11. Enabling Science: DEISA users communities National users communities have accounts on a given site and do not « naturally » see the whole DEISA environment. Promotion of DEISA users is done via the Extreme Computing Initiative. Europ. call for proposals for grand challenge simulations every year in May since 2005. About 50 grand challenge projects supported each year since 2005. Full information about Extreme Computing projects and reports from terminated projects can be found on the DEISA Web server: www.deisa.eu The Extreme Computing Initiative is the current DEISA service provisioning model. HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 11 RI-222919
  • 12. One Example: Joint Research Activity “Life Sciences” Promoting parallel apps in the life science community Running big simulations on DEISA infrastructure that couldn’t be done locally Providing ease of access to resources Application support for life science portal DEISA Life Science Portal HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 12 RI-222919
  • 13. DEISA Life Science Portal based on NICE EnginFrame HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 13 RI-222919
  • 14. Portal Deployment Computing Nodes Submission Host EF Agent / Site Application Scheduler Architecture Architecture SSH Plugin i586 BSC NAMD IBM Power PC Linux SLURM/MOAB SSH Suse Linux 10 IDRIS BLAST IBM Power6 AIX IBM Power6 AIX LoadLeveler Agent Intel Itanium 2 LRZ RAxML SGI Altix Linux PBSPRO Agent Suse Linux 9 IBM Power 5 CINECA BLAST IBM Power5 AIX LoadLeveler SSH AIX 5.3 IBM Power 4+ FZJ BLAST IBM Power4+ AIX LoadLeveler SSH AIX 5.3 IBM Power 4 RZG BLAST IBM Power4 AIX LoadLeveler Agent AIX 5.3 Power 5 EPCC BLAST IBM Power5 AIX LoadLeveler SSH AIX 5.3 CRAY XT4 Opteron CSC NAMD PBSPRO Agent Opteron Unicos Suse Linux 9 IBM Power 5 SARA NAMD IBM Power 5 Linux LoadLeveler Agent Suse Linux 9 HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 14 RI-222919
  • 15. DEISA Extreme Computing Initiative DECI Application Enabling and Optimizations • Scaling of parallel programs for efficient usage on thousands of processor-cores challenging, necessary task for state-of-the-art supercomputers • Support for data intensive applications • Design, deployment and optimization of workflow applications to chain several compute tasks (simulation, pre- and post-processing steps) • Coupled applications important e.g. for climate system modelling with separate components for ocean, atmosphere, sea ice, soil, atmospheric chemistry and aerosols • Adaptation to the DEISA infrastructure • Optmizations, also architecture dependent optimizations • Determination of best suited architecture HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 15 RI-222919
  • 16. DEISA Extreme Computing Initiative Annual Calls for Proposals for challenging supercomputing projects from all areas of science DECI call 2005 > 50 proposals, 12 European countries involved, co-investigators from US > 30 mio cpu-h requested 28 projects selected, 12 mio cpu-h awarded DECI call 2006 > 40 proposals, 12 European countries involved, co-investigators from N + S America, Asia (US, CA, AR, ISRAEL) ~ 30 mio cpu-h requested 23 projects selected, 12 mio cpu-h awarded DECI call 2007 > 60 proposals, 14 European countries involved, co-investigators from N + S America, Asia, Australia (US, CA, BR, AR, ISRAEL, AUS) > 70 mio cpu-h requested 45 projects selected, ~ 30 mio cpu-h awarded DECI call 2008 - 66 proposals, 14 European countries, USA (7), China, Japan involved - 134 mio cpu-h requested HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 16 RI-222919
  • 17. Extreme Computing Applications for HPC, Supercomputing, Capacity Computing www.deisa.eu RI-222919
  • 18. Aerodynamic Shape Optimization DOE plan Geometry modification Volume meshing Optimizer CFD simulation • 4 parameters to be optimized Result analysis • cubic face centered DOE • 25 cases+16 extra cases for error estim. Optimal • polynomial response function • 70 hours wall clock time on 64 cpus solution Each of these steps need to be fully automated and controlled by the optimizer HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 18 RI-222919
  • 19. Combustion / Radiation – Study the impact of radiative N processes MPI t + m.dt Pthreads (OpenMP) Species, T heat transfer (RHT) on the AVBP RAYON combustion process (2D) (combustion) Radiative E (RHT) Yi, T, V n.dt t+ – Couple combustion (AVBP), the RHT (Rayon) codes and the pollutant formation AVBP (AVBP) (pollutant) N processes MPI – Parallelization of the Rayon code and improvement of the Temperature field Radiative heat transfer coupling part WITH WITHOUT – Load balancing issue – 3D extension proposed to DECI and accepted HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 19 RI-222919
  • 20. Climate Forecast Climate research moves towards new levels of complexity: Stepping from Climate (=Atmosphere+Ocean) to Earth System Modelling Earth system model wishlist: Higher spatial and temporal resolution Quality: Improved subsystem models Atmospheric chemistry (ozone, sulfates,..) Bio-geochemistry (Carbon cycle, ecosystem dynamics,..) Increased Computational demand factor: O(1000 -10000) HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 20 RI-222919
  • 21. Climate Research Statistics of Climate Variability Project to study climate trends, each 50 TB output data H. Dijkstra, U Utrecht, and W. Hazeleger, KNMI HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 21 RI-222919
  • 22. Environmental Application – Study the impact of water cycles of the hydrological and vegetation models on climate models – Coupling area in West Africa – Best performances with a Hydrological Mesh Meteorological Atmosphere Vegetation Hydrological vector and scalar platform Mesh MAR SVAT_M – Improve extensibility of the architecture and the Sirba Sirba coupling part SVAT_H_1 ABC Ouémé Oué Ouémé Oué – AMMA project, PhD thesis, SVAT_H_2 TopModel 2 publ. and 2 comms. HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 22 RI-222919
  • 23. Materials Science First-principles statistical mechanics for molecular switches at surfaces (MolSwitch) Azobenzene on copper, silver and gold surfaces Controlled reversible switching should be possible on Ag surfaces Courtesy: K. Reuter, FHI HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 23 RI-222919
  • 24. Polymer Research Cover story of Nature - May 24, 2007 Curvy membranes make proteins attractive For almost two decades, physicists have been on the track of membrane mediated interactions. Simulations in DEISA have now revealed that curvy membranes make proteins attractive Nature 447 (2007), 461-464 a) proteins (red) adhere on a membrane (blue/yellow) and locally bend it; b) this triggers a growing invagination. c) cross-section through an almost complete vesicle B. J. Reynwar et al.: Aggregation and vesiculation of membrane proteins by curvature mediated interactions, NATURE Vol 447|24 May 2007| doi:10.1038/nature05840 HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 24 RI-222919
  • 25. Cosmology Project – Study galaxy formation in Hydrodynamics cosmology (Baryonic gas) s Ab r e a as fo m un ch – Physics / modules: as phi da spe g nc cie Gravitation, Hydrodynamics, e Chemistry Gravity Chemistry – Best performance on (Dark Matter) (Baryonic gas) heterogeneous platforms – Load balancing issue and improvement of the coupling part – Proposed to DECI a) b) HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 25 RI-222919
  • 26. PLASMA Physics Theory Support of ITER Experiment HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 26 RI-222919
  • 27. PLASMA Physics Theory Support of ITER Experiment HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 27 RI-222919
  • 28. Plasma Physics Simulations Scaling Enabling and hyperscaling plasma physics codes 4096 processors 8192 processors 2048 processors norm. speed-up 16 4 relative speed-up 2000 ITER 2005 14 JET GEM 2007 1500 12 2006 AUG 3 speedup 10 8 1000 6 ORB5 2 4 500 TORB 2 0 1 0 0 5000 10000 15000 20000 500 1000 1500 2000 0 500 1000 1500 2000 number of processors number of cores number of nodes 32768 processors 16384 processors 120% 8 7 100% parallel efficiency 6 2007 80% 5 speedup 4 60% 2007 3 2 GENE 40% 1 20% GENE 0 0 2000 4000 6000 8000 10000 0% number of processors 1 10 100 1000 10000 100000 number of processors HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 28 RI-222919
  • 29. Read more in: Thank You! Gentzsch@rzg.mpg.de HPCC, March 24-26 2009 Wolfgang Gentzsch, DEISA 29 RI-222919