ParaView In Action
This page captures blurbs about real-world usage of ParaView. The purpose is to compile a list of success stories with engaging images that can be used in things like web pages, posters, and elevator speeches.
Verification of ALEGRA Simulations
Computer simulation is a key component in modern scientific analysis. Simulations provide more control over variables and more detail in their outputs than their experimental counterparts while allowing for more runs at a fraction of the price. However, simulation results are worthless without verification, which ensures that a simulation can reliably compute the results of a simulation with a known analytical solution.
Verification becomes more difficult with the growth of size and complexity in the simulations. Analysts at Sandia National Laboratories are leveraging the python scripting, scalable parallel processing, and multiview technologies of ParaView to simplify the verification of ALEGRA simulations for High Energy Density Physics (HEDP). The embedded python interpreter allows users to easily build custom parallel computations for analytical solutions, volume integrals, and error metrics over large meshes. Client side scripting also facilitates automated post-processing verification and ParaView’s multiview capability facilitates comparison of results. The visualization shown here displays the computed error metrics for a 13 million-cell unstructured data set.
Distance Visualization for Terascale Data
Visualizing the worlds largest simulations introduces a multitude of pragmatic issues. Moving such large quantities of data is infeasible. The simulation hardware is often unique, complex, and built without visualization or post processing in mind. Furthermore, the hardware resources may be located far away and without the high speed connections expected in a local network. Such was the case for a set of analysts and Sandia National Laboratory studying high energy density physics (HEDP). HEDP simulations ran on the ASC Purple supercomputer located in Lawrence Livermore National Laboratory, over 850 miles (1300 km) from the analysts' office. Any communication between these two sites required the use of SecureNET, an encrypted communication channel with moderate communication speeds (45-600 Mbps). ParaView provides the remote analysis capability our scientists need. ParaView is deployed on ASC Purple "out-of-the-box" despite the complexity of hardware, and ParaView's quality parallel rendering and image delivery mechanism make remotely interacting with the data simple and effective.
Golevka Asteroid
In this CTH shock physics simulation, a 10 megaton explosion is detonated at the center of the Golevka asteroid, which has an approximate volume of 210 million cubic meters (500m x 600m x 700m). The mesh the simulation uses has a 1 meter resolution consistent over the 1 cubic kilometer mesh, resulting in over 1 billion cells. The remarkable resolution of this simulation provides realism in crack propagation not seen in lower-resolution models. The simulation was run for 15 hours on 7200 nodes of Sandia National Laboratories' Red Storm supercomputer.
The visualization, here showing a cut through the center of the asteroid with velocity magnitude colored on the cut surface, is enabled via ParaView's scalable visualization code running in parallel on 128 visualization nodes.
Polar Vortex Breakdown
This terabyte SEAM Climate Modeling simulation models the breakdown of the polar vortex, a circumpolar jet that traps polar air at high latitudes, creating conditions favorable for ozone depletion. The breakdown of the vortex, which occurs once or twice a year in the polar wintertime stratosphere, can transport ozone-depleting polar air well into the mid latitudes. The 1 billion cell structured mesh reveals fine details of the weather pattern not represented in previous simulations.
The visualization is enabled via ParaView's scalable visualization and rendering code running on over 128 visualization nodes delivered to the analyst's desktop computer at interactive rates.
Cross Wind Fire
This 150 million degree-of-freedom, loosely coupled SIERRA/Fuego/Syrinx/Calore simulation models objects-in-crosswind fire. It features three regions: one fluids and participating media radiation region and two conducting regions. The simulation was run on 5000 nodes of Sandia National Laboratories' Red Storm supercomputer and is part of a qualification test plan for system testing to be conducted at the new Thermal Test Complex Cross Wind Facility.
The visualization, showing the temperature of the gasses, showcases the unique parallel unstructured volume rendering capabilities feature of ParaView contributed by Sandia National Laboratories.
Parallel Rendering
Sandia National Laboratories, in conjunction with NVIDIA Corporation and Kitware, Inc., have established ParaView as the world leader in parallel rendering performance. In a November 2005 press release, Sandia demonstrated ParaView delivering over 8 billion polygons per second rendered on the Red RoSE visualization cluster and delivered to an analyst's desktop. ParaView's record breaking rendering speeds are made possible with Sandia's parallel rendering library (IceT) and Sandia's image compression system (SQUIRT).
3D Rayleigh-Benard convection problem
This problem was used to investigate EdgeCFD's code performance in a large scale simulation. EdgeCFD is an implicit edge-based coupled fluid flow and transport solver for solving large scale problems in modern clusters, supporting stabilized and variational multiscale finite element formulations IJNMF2007. The benchmark corresponds to a rectangular 3D domain of aspect ratio 4:1:1 aligned with the Cartesian axes and subjected to a temperature gradient. The simulation was made on a 501×125×125 mesh, resulting in 39,140,625 tetrahedral elements. The figure shows the convective rolls obtained at Rayleigh number Ra=30,000 and Prandt number Pr=0.71. This solution was obtained on 128 cores of a SGI Altix ICE 8200 cluster installed at High Performance Computing Center NACAD of Federal University of Rio de Janeiro UFRJ. Every time step employed an Inexact Newton method and two nonlinear systems of equations, for flow and temperature, respectively with 31M and 7.8M equations. Time step solutions were stored using the Xdmf file format in a temporal collection of geometry collections scheme, summarizing 1936 files for 15/2955 (stored/total). The simulation took about 170 hours (1 week approximately) and the post-processing was done using ParaView in a remote-client x server offscreen rendering scheme.
Acknowledgements
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000.
