11
High Performance Computing: High Performance Computing: Applications in Science and Applications in Science and Engineering Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH IITK REACH Symposia 2010 Symposia 2010 10-12 October’10 10-12 October’10 Amalendu Chandra (Convener) Amalendu Chandra (Convener) Kalyanmoy Deb Kalyanmoy Deb

High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

Embed Size (px)

Citation preview

Page 1: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

High Performance Computing: High Performance Computing: Applications in Science and Applications in Science and

EngineeringEngineering

REACH Symposium on HPC 10 October 2010

1

IITK REACH IITK REACH Symposia 2010Symposia 201010-12 October’10 10-12 October’10

Amalendu Chandra (Convener) Amalendu Chandra (Convener) Kalyanmoy DebKalyanmoy Deb

Page 2: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

2

High Performance Computing (HPC)High Performance Computing (HPC)Usage of computers for solving complex Usage of computers for solving complex computing problemscomputing problems

Nanotechnology, molecular modeling, biologyNanotechnology, molecular modeling, biology3D modeling and simulation, optimization3D modeling and simulation, optimizationCFD, weather modelingCFD, weather modelingGaming, Data mining, etc.Gaming, Data mining, etc.

Usually requiring TeraFLOP of computations Usually requiring TeraFLOP of computations (Trillion floating point operations per second)(Trillion floating point operations per second)Indispensible companion for higher education Indispensible companion for higher education and research todayand research today

Slide rules for 60-70s, calculators for 80s, Slide rules for 60-70s, calculators for 80s, miniframes and mainframes for 80-90sminiframes and mainframes for 80-90s

Direct correlation between computing power Direct correlation between computing power and extent/quality of researchand extent/quality of research

Page 3: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

3

Different Computing ParadigmsDifferent Computing ParadigmsHigh Performance Computing (HPC)High Performance Computing (HPC)High Performance Technical Computing (HPTC)High Performance Technical Computing (HPTC)High Throughput Computing (HTC)High Throughput Computing (HTC)Many-task Computing (MTC)Many-task Computing (MTC)MetacomputingMetacomputingSupercomputingSupercomputingGrid ComputingGrid ComputingCloud ComputingCloud ComputingParallel ComputingParallel ComputingDistributed ComputingDistributed ComputingHeterogeneous ComputingHeterogeneous ComputingFuturistic: Futuristic: Quantum ComputingQuantum Computing

Page 4: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

4

Top FIVE Performers (June 2010)Top FIVE Performers (June 2010)On Linpack benchmarkOn Linpack benchmark

Cray XT5 Cray XT5 (Jaguar), 1759 Tflops, ORNL, USA, (Jaguar), 1759 Tflops, ORNL, USA, 2009 2009 Dawning TC3600 Dawning TC3600 (Nebulae), 1271 Tflops, NSCS, (Nebulae), 1271 Tflops, NSCS, Shenzhen, China, 2010Shenzhen, China, 2010IBM BladeCenter IBM BladeCenter (Roadrunner), 1042 Tflops, (Roadrunner), 1042 Tflops, Los Alamos, USA, 2009Los Alamos, USA, 2009Cray XT5 Cray XT5 (Kraken), 871 Tflops, NICS, USA, 2009(Kraken), 871 Tflops, NICS, USA, 2009IBM BlueGene IBM BlueGene (Jugene), 825 Tflops, Julich, (Jugene), 825 Tflops, Julich, Germany, 2009Germany, 2009

Page 5: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

5

PetaFlop to ExaFlop ProjectionsPetaFlop to ExaFlop ProjectionsPetaFlop: 1000 TeraFlopsPetaFlop: 1000 TeraFlopsCurrent Best: 1.76 PetaFlopCurrent Best: 1.76 PetaFlopMoore’s Law: Speed doubling every 14 mths, 2009 Moore’s Law: Speed doubling every 14 mths, 2009 machine is 18,000 times faster than that in 1993machine is 18,000 times faster than that in 19932011-2012: 20 PetaFlop machine by IBM & Office of 2011-2012: 20 PetaFlop machine by IBM & Office of Science, USA (design confirmed)Science, USA (design confirmed)2015: 100-250 PetaFlop machine (projected)2015: 100-250 PetaFlop machine (projected)2018: ExaFlop machine (1000 PetaFlops) 2018: ExaFlop machine (1000 PetaFlops) (projected)(projected)Might see in our lifetime ZettaFlop, YottaFlop, Might see in our lifetime ZettaFlop, YottaFlop, XonaFlop, WekaFlop or VundaFlop machinesXonaFlop, WekaFlop or VundaFlop machinesOr may there be a different type of computing Or may there be a different type of computing paradigm!paradigm!

Page 6: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

6

HPC Facilities in IndiaHPC Facilities in IndiaIndia (in Top500, June’10, India (in Top500, June’10, http://www.top500.orghttp://www.top500.org) )

Rank 33: Computational Research Laboratories, TATA Rank 33: Computational Research Laboratories, TATA SONS, SONS, 132.8 Tflops132.8 Tflops, HP Cluster, 2008, HP Cluster, 2008Rank 93: Indian Institute of Tropical Meteorology. Rank 93: Indian Institute of Tropical Meteorology. 55.1 55.1 TflopsTflops, IBM Power 575, 2010, IBM Power 575, 2010Rank 182: CDAC, Rank 182: CDAC, 38.1 38.1 Tflops, Param Yuva, 2008Tflops, Param Yuva, 2008Rank 369: Rank 369: IIT KanpurIIT Kanpur, , 29.0129.01 Tflops, HP Cluster, 2010 Tflops, HP Cluster, 2010Rank 389: IT Services Provider, Rank 389: IT Services Provider, 28.3628.36 Tflops, HP Tflops, HP Cluster, 2009Cluster, 2009

IITK is the only IIT/IISc. featuring in Top500IITK is the only IIT/IISc. featuring in Top500China: Rank 2, 2 in Top10 and 24 in Top500China: Rank 2, 2 in Top10 and 24 in Top500HiPC: An Intl. conference in India since 1995HiPC: An Intl. conference in India since 1995

Page 7: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

7

Usage of HPCUsage of HPCTask can be small or largeTask can be small or largeTask may involve an uniprocessor or Task may involve an uniprocessor or multiprocessorsmultiprocessorsTask may be compute-intensive, data-intensive, Task may be compute-intensive, data-intensive, or communication-intensive, or a combinationor communication-intensive, or a combinationTask may be static or dynamicTask may be static or dynamicTask may be homogeneous or heterogeneousTask may be homogeneous or heterogeneousTask may be loosely coupled or tightly coupledTask may be loosely coupled or tightly coupled

Page 8: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

8

HPC ResearchHPC ResearchNew applicationsNew applications

Venturing to solve problems which were not Venturing to solve problems which were not possible to solve earlierpossible to solve earlier

New algorithms/methodologies for problem New algorithms/methodologies for problem solvingsolvingSolving existing problems more efficientlySolving existing problems more efficientlyNew implementations – compilers, code New implementations – compilers, code parallelizationparallelizationNew hardware developmentNew hardware development

Faster, more data-handling, faster communicationsFaster, more data-handling, faster communications

““Solution is of the computer model, need not be Solution is of the computer model, need not be of the real system” of the real system”

Page 9: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

9

Symposium on HPCSymposium on HPCSanjay Mittal (IITK): Sanjay Mittal (IITK): Using HPC for understanding of Using HPC for understanding of Fluid FlowsFluid FlowsS. Sherlekar (INTEL): S. Sherlekar (INTEL): An An ApplicationsApplications Perspective of Perspective of High-Performance Computing High-Performance Computing R. Sankararamakrishnan (IITK): R. Sankararamakrishnan (IITK): From sequence From sequence analysis to simulations: Applications of high-analysis to simulations: Applications of high-performance computing in modernperformance computing in modern biology biologySanjeev Aggarwal (IITK): Sanjeev Aggarwal (IITK): Automatic Automatic ParallelizationParallelization of Programsof ProgramsPrashant Goswami (C-MMACS): Prashant Goswami (C-MMACS): Will HPC ever meet Will HPC ever meet the demands of the demands of weather and climate forecastingweather and climate forecasting??Amalendu Chandra (IITK): Amalendu Chandra (IITK): Molecular simulations Molecular simulations and HPC@IITKand HPC@IITK30 Posters on various 30 Posters on various applications (IITK)applications (IITK)

Page 10: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

Acknowledgements

REACH Symposium on HPC 10 October 2010

10

Patron, Prof. S. G. DhandePatron, Prof. S. G. DhandeGolden Jubilee CommitteeGolden Jubilee CommitteeV. Chandrasekhar and his REACH organising V. Chandrasekhar and his REACH organising team and support staffteam and support staffK. Muralidhar and Manindra Agrawal, REACH K. Muralidhar and Manindra Agrawal, REACH Advisory CommitteeAdvisory CommitteeAll the invited speakersAll the invited speakersAdditional thanks to IITK invited speakers for Additional thanks to IITK invited speakers for their time for the pre-presentationstheir time for the pre-presentationsAll the poster authors and presentersAll the poster authors and presentersSpecial thanks to Prof. Gautam Biswas …Special thanks to Prof. Gautam Biswas …

Page 11: High Performance Computing: Applications in Science and Engineering REACH Symposium on HPC 10 October 2010 1 IITK REACH Symposia 2010 10-12 October’10

REACH Symposium on HPC 10 October 2010

11

Chair of theChair of theSessionSession

Prof. Gautam BiswasProf. Gautam BiswasDirector, Central Mechanical Engineering ResearchDirector, Central Mechanical Engineering ResearchInstitute (CMERI) Durgapur andInstitute (CMERI) Durgapur andProfessor, Department of Mechanical Engineering, Professor, Department of Mechanical Engineering, IIT KanpurIIT Kanpur