1 / 22

John Clyne & Alan Norton Scientific Computing Division National Center for Atmospheric Research

VAPOR Visualization and Analysis Platform for Ocean, atmosphere, and solar Research SC06 Ultra-Scale Visualization Workshop. John Clyne & Alan Norton Scientific Computing Division National Center for Atmospheric Research Boulder, CO USA.

Download Presentation

John Clyne & Alan Norton Scientific Computing Division National Center for Atmospheric Research

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. VAPORVisualization and Analysis Platform for Ocean, atmosphere, and solar ResearchSC06 Ultra-Scale Visualization Workshop John Clyne & Alan Norton Scientific Computing Division National Center for Atmospheric Research Boulder, CO USA This work is funded in part through a U.S. National Science Foundation, Information Technology Research program grant clyne@ncar.ucar.edu

  2. Problem • Numerical simulations in the earth sciences have reached such extraordinary sizes that researchers can no longer effectively extract insight from their simulation outputs. • Result: loss of scientific productivity!!! [Numerical] models that can currently be run on typical supercomputing platforms produce data in amounts that make storage expensive, movement cumbersome, visualization difficult, and detailed analysis impossible.  The result is a significantly reduced scientific return from the nation's largest computational efforts. Mark Rast University of Colorado, LASP clyne@ncar.ucar.edu

  3. Dichotomy of simulation and analysis needs and resources in today’s HPC environments clyne@ncar.ucar.edu

  4. A sampling of various technology performance curves • Not all technologies advance at same rate • Impact of parallelization not shown clyne@ncar.ucar.edu

  5. Communication limits for volume rendering assuming theoretical peak performance • Table shows limits expressed as frames per second imposed by communication alone • Assumes only 8-bit data quantities clyne@ncar.ucar.edu

  6. Visual data browsing Refine Coarsen Quantitative analysis Data manipulation Visualization and Analysis Platform for Ocean, atmosphere, and solar Research (VAPOR) Key components • Domain specific application focus: simulated earth sciences fluid flow • Coupled Visualization and quantitative data interrogation and manipulation capabilities • Multiresolution enabled terascale data exploration on the desktop Combination of visualization with multiresolution data representation that provide sufficient data reduction to enable interactive work on terascale data from a desktop clyne@ncar.ucar.edu

  7. Fluid flow in the geosciences • E.g. Numerically simulated turbulence • Cartesian grids (usually) • 5123 to 10243 • Up to 40963 “hero” calculations • 5 to 8 variables • Temperature & Pressure • Velocity field components • Magnetic field components (MHD calculations) • Hundreds of time steps saved • Terabytes of data per experiment • Numerical “experiments” • Substantial analysis requirements Yannick Ponty, CNRS 2006 clyne@ncar.ucar.edu

  8. Key Component (1) : Domain specific support Only limited support for: • Grid & data types • Cartesian grids, stretched and uniform sampling • AMR grids • Scalar and vector quantities • Visualization algorithms • Volume rendering, flow visualization, cutting planes/probe • Misc. • Publication quality graphics • Filters • File formats (one!) • Extensive support for: • Time varying data • Uniform as well as non-uniform sampling • Missing time steps • Quantitative investigation • Mathematical operators and data manipulators • Science driven specialized features Keep it simple! Keep it focused! Make it scientist friendly! clyne@ncar.ucar.edu

  9. Interactive exploration of time varying data • Reduce bandwidth requirements • Regions of interest • Multiresolution • Caching clyne@ncar.ucar.edu

  10. VAPOR Interactive visual browsing Future??? IDL Data manipulation & analysis VAPOR Data Collection Multi-resolution access and rapid sub-region extraction Disk Array Key Component (2) : Coupled visualization, quantitative analysis and manipulation capabilities • IDL - array based 4GL for scientific data processing • Thousands of mathematical functions • Basic 2D plotting • Array manipulation clyne@ncar.ucar.edu

  11. Wavelet transformed data Two parameter linear function decomposition Hierarchical data representation Invertible and lossless Numerically efficient (O(n)) forward and inverse transform No additional storage cost Enable speed/quality tradeoffs Key component (3) : Multiresolution data access 504x504x2048 Full 252x252x1024 1/8 126x126x512 1/64 63x63x256 1/512 clyne@ncar.ucar.edu

  12. Visual comparison of a 5123 compressible convection simulation M. Rast, 2002 1283 coarsened 5123 native clyne@ncar.ucar.edu

  13. Performance of forward and inverse Haar wavelet transform • Data • Scalar • Single precision • System • Linux RHEL 3.0 • 2 x Intel 3.4 GHz Xeon EMT64 • 8 GBs RAM • 1Gb/sec Fibre Channel storage Gains in microprocessor technology enable transforms at very low cost clyne@ncar.ucar.edu

  14. VAPOR Demo clyne@ncar.ucar.edu

  15. Summary • VAPOR is a domain-specific platform for analysis, not a general purpose visualization tool • Target users: fluid flow researchers in earth sciences • Limited value for medical, oil & gas, aerospace, etc. • Desktop data exploration of terabyte data possible • Visualization enables rapid ROI identification • Multiresolution enables speed/quality tradeoffs clyne@ncar.ucar.edu

  16. Steering Committee Nic Brummell - CU Yuhong Fan - NCAR, HAO Aimé Fournier – NCAR, IMAGe Pablo Mininni, NCAR, IMAGe Aake Nordlund, University of Copenhagen Helene Politano - Observatoire de la Cote d'Azur Yannick Ponty - Observatoire de la Cote d'Azur Annick Pouquet - NCAR, ESSL Mark Rast - CU Duane Rosenberg - NCAR, IMAGe Matthias Rempel - NCAR, HAO Geoff Vasil, CU Developers Alan Norton – NCAR, SCD John Clyne – NCAR, SCD Kenny Gruchalla - CU Research Collaborators Kwan-Liu Ma, U.C. Davis Hiroshi Akiba, U.C. Davis Han-Wei Shen, Ohio State Liya Li, Ohio State Systems Support Joey Mendoza, NCAR, SCD Acknowledgements clyne@ncar.ucar.edu

  17. Questions??? www.vapor.ucar.edu clyne@ncar.ucar.edu

  18. Inverse Haar transform with 1/8th volume subregion extraction • System • Linux RHEL 3.0 • 2 x Intel 3.4 GHz Xeon EM64 • 8 GBs RAM • 1Gb/sec Fibre Channel storage • Data • Scalar • Single precision Data blocking permits rapid subregion extraction clyne@ncar.ucar.edu

  19. The Lifting Method of wavelet construction in the spatial domain[Sweldens, 95] A signal λjconsisting of 2j samples Split signal into even (λ) and odd (γ) coefficients. λ will contain low frequency information, γ will contain high frequency information. 1) Split: 2) Predict: Local correlation permits prediction of odd samples by even using a prediction operator, P. Capture difference between prediction and actual coefficient value. 3) Update: Update λ coefficients to preserve a property (e.g. mean) of original signal. Update Split Predict Transform 2 Transform j Transform 1 clyne@ncar.ucar.edu

  20. Haar operators 1 7 3 1 6 0 9 5 γ2,k 6 -2 -6 -4 4 2 3 7 -2 4 3 5 2 4 Example: Lifting Method with the Haar Wavelet λ3,k λ2,k γ1,k λ1,k λ0,k γ0,k clyne@ncar.ucar.edu

  21. NCAR Historical Estimated Sustained GFLOPS (Batch Production Systems) clyne@ncar.ucar.edu

  22. NCAR Historical Estimated Sustained GFLOPS (Interactive Production Systems) • Current NCAR visualization and analysis resources • ~32 processors • 8 nodes (6 with gfx) • ~100 TB on-line storage • ~800 MBs/sec aggregate storage bandwidth • ~100 users (99 of which will not leave office) clyne@ncar.ucar.edu

More Related