The only thing as daunting as the computational problem of simulating a tornado may be standing in the path of an actual one.
Leigh Orf is one of the leading computational scientists studying tornadoes. Based at the Cooperative Institute for Meteorological Satellite Systems, at the University of Wisconsin, Madison, he researches the conditions that trigger the highest category of tornadoes. His simulations generate terabytes of data.
Understanding the spatial relationships of the elements that lead to a tornado requires volume visualizations. Due to the huge data sizes Orf works with, he needs an iterative process consisting of multiple layers of isosurfaces to analyze various attributes including pressure, wind speed, and temperature.
Given the highly turbulent nature of tornadoes, selecting the right isosurface values can be a time-consuming task, fraught with trial and error. Exploration requires looking at the data from different perspectives or changing visualization parameters, like opacity. If modifying parameters results in lengthy recalculations, the flow of the discovery process is interrupted and lost.
A Picture Is Worth a Terabyte of Data
With high-end GPUs delivering supercomputer performance on a single chip, raw processing power is no longer the most precious resource in high performance computing. The main challenge is converting the terabytes and petabytes of numbers generated by simulations like Orf’s into scientific insight.
Originally designed for the data visualization challenges encountered in the oil and gas industry, NVIDIA IndeX is a volume visualization tool that delivers the highest interactive performance on large datasets by distributing workloads across GPU-accelerated clusters.
We’ve recently collaborated with Kitware to bring IndeX’s capabilities to the broad scientific computing community as a plug-in for ParaView, a widely used open source data analysis and visualization application.
IndeX for ParaView lets researchers quickly visualize and interact with their data — at over 10 frames per second for terabyte and larger datasets — facilitating an intuitive understanding of otherwise overwhelming data.
And it does so without requiring them to copy huge datasets to local file systems. Instead of implementing compromised workarounds, scientists can use IndeX to get the highest visualization performance by distributing data across multiple nodes within a GPU-accelerated cluster.
IndeX also simplifies workflows by allowing users to take advantage of the ParaView interface. Scientists can focus on their research and accelerate discoveries rather than spending time learning a new tool.
IndeX for ParaView can be run inside a workstation or on a GPU-accelerated cluster. The workstation plug-in is developed for both Windows and Linux OS and is available at no cost. The cluster edition can be purchased with a commercial license.
Learn more about the NVIDIA IndeX for ParaView plug-in at www.nvidia.com/index.