A Software Model for Visualization of Large Unsteady 3-D CFD Results

Al Globus, Computer Sciences Corporation
NASA Ames Research Center


The largest unsteady 3-D computational fluid dynamics (CFD) results produced today exceed 100 gigabytes. Visualizing these large data sets stresses computational resources, requiring users and visualization systems to make trade-offs between time, space, and flexibility. Current CFD visualization systems make trade-offs which severely limit visualization of the largest unsteady 3-D CFD results. We propose a software model that breaks visualization processing into several stages: generation of solutions, extracts, graphic objects, scenes, and movies (explained below). Discarding data from early stages saves storage. Postponing decisions to later stages improves flexibility and speed. We also propose storing these data in a persistent object database in order to separate the logical and physical representations of the data and provide a consistent, clean interface between visualization software and data from all stages. Meta-data, such as performance information, may also be stored to support intelligent time vs. space vs. flexibility trade-offs. Ultimately, more optimal trade-offs expand the visualizations possible within constant hardware and wall clock time limits. This paper will describe, explain, and motivate the proposed model. In principle, the model should be of value in disciplines other than CFD.