visualization and data analysis using visit
play

Visualization and Data Analysis using VisIt - In Situ Visualization - PowerPoint PPT Presentation

Mitglied der Helmholtz-Gemeinschaft Visualization and Data Analysis using VisIt - In Situ Visualization - Jens Henrik Gbbert 1 , Herwig Zilken 1 1 Jlich Supercomputing Centre, Forschungszentrum Jlich GmbH, Germany Application Support,


  1. Mitglied der Helmholtz-Gemeinschaft Visualization and Data Analysis using VisIt - In Situ Visualization - Jens Henrik Göbbert 1 , Herwig Zilken 1 1 Jülich Supercomputing Centre, Forschungszentrum Jülich GmbH, Germany Application Support, Cross-Sectional- Team „ Visualization “ {j.goebbert, h.zilken}@fz-juelich.de July 8., 2016 VisIt Workshop VisIt Workshop Folie 1

  2. Motivation why not stick to the established workflow Application CIAO on JUQUEEN (specific setup) Excellent scaling of solver ■ File writing dumping output is bottleneck at scale ■ dump simulation data with high temporal resolution not possible July 8., 2016 VisIt Workshop Folie 2

  3. In Situ Visualization visualization for large-scale simulations … post visualization Visualization of simulation data in a post processing step requires to write and then read files, paying for I/O twice in different application. … in situ visualization  visualization capabilities inside the simulation  avoid I/O  compression method July 8., 2016 VisIt Workshop Folie 3

  4. Motivation lowering the barriers to in-situ visualization Major barriers to in situ visualization are … … first,  the individual implementation-, optimization- and coupling-costs to integrate the needed functionality to each simulation code and setup can often not be justified. … second,  the usage of in situ visualization requires much training for scientists who's research work in general does not focus on visualization in the first place. Lowering the barriers to in-situ visualization July 8., 2016 VisIt Workshop Folie 4

  5. Applications coupled with VisIt/Libsim … psOpen … CIAO … ZFS  flow solver (DNS)  flow solver (LES, DNS)  flow solver (LES, DNS)  highly resolved turbulence  multiphysics, multiscale  multiphysics, multiscale  pseudo-spectral approach  structured finite difference method  finite volume method  level-set for surface tracking  lattice-Boltzmann method  level-set/volume-of-fluid interface  discontinuous Galerkin method  Institute for Combustion Technology  Lagrange particle solver  level-set for surface tracking RWTH Aachen University, Germany  tabulated/finite rate chemistry  Lagrangian particle solver  Chair of Num. Thermo-Fluid Dyn.  overset mesh refinement TU Freiberg, Germany  Institute of Aerodynamics Aachen  moving meshes RWTH Aachen University, Germany  Fortran90 + MPI + OpenMP  Institute for Combustion Technology  RWTH Aachen University, Germany C++11 + MPI + OpenMP + GPU  JUQUEEN BigWeek participant  Fortran90 + MPI  JUQUEEN BigWeek participant July 8., 2016 VisIt Workshop Folie 5

  6. Staged In Situ „small“ compute node „big“ compute node GPU HPC file system interactive visualization node HPC interconnect GPU • „ small “ compute nodes für die Simulation • „ big “ compute nodes für die nachgelagerte Verarbeitung • Visualisierung/Analyse der Rohdaten • unmittelbares Rendering auf „ big “ nodes mit OpenGL-fähigen GPUs Pro: exklusive Hardwareresourcen für Simulation und nachgelagerte Verarbeitung Contra: Kopieren der Daten zwingend notwendig July 8., 2016 VisIt Workshop Folie 6

  7. On-Node In-Situ „small“ compute node HPC file system interactive visualization node HPC interconnect GPU • „ small “ compute nodes • Simulation und Visualisierung/Analyse der Rohdaten Pro: Kopieren von Daten zwischen Simulation und Datenverarbeitung entfällt Contra: keine exklusiven Hardwareresourcen für Simulation und Verarbeitung July 8., 2016 VisIt Workshop Folie 7

  8. Coupling of Simulation and Visualization/Analysis „ no coupling “ Simulation Visualization „ tightly-coupling “ Simulation Visualization „ loosly-coupling “ In-Situ Simulation Visualization Framework 8 July 8., 2016 VisIt Workshop Folie 8

  9. Tightly-Coupled In-Situ Visualization 9 July 8., 2016 VisIt Workshop Folie 9

  10. Scalable Rendering Mode VisIt – parallelized client-server architecture Local Components Parallel Cluster VisIt Data Data Engine Plugin connection network MPI VisIt Data Data Engine Plugin VisIt Data Data Plugin Engine VisIt (Files or Simulation) Viewer Data Flow Network Filter VisIt VisIt Java Python Filter Clients GUI CLI Clients Filter July 8., 2016 VisIt Workshop Folie 10

  11. JUSITU coupling simulation code to in-situ visualization VisIt/ Libsim simulation ParaView/ Catalyst sim. developer vis. developer  one data adapter for each visualization tool  need for reinventing infrastructure code for each simulation to couple  responsibility for coupling code by sim. developer  changes on the visualization software side only possible in close cooperation with sim. developers July 8., 2016 VisIt Workshop Folie 11

  12. JUSITU coupling simulation code to in-situ visualization VisIt/ Libsim simulation JUSITU ParaView/ Catalyst sim. developer vis. developer  light-weighted, flexible and easy-to-use coupling library  covers the complexity and numerous options of in-situ visualization  simplifies communication between sim. developer and vis. developer  written in C++ / MPI July 8., 2016 VisIt Workshop Folie 12

  13. In Situ Visualization with VisIt/Libsim July 8., 2016 VisIt Workshop Folie 13

Recommend


More recommend