PARALLEL simulations now play an important role in all ar-eas of science and engineering (fluid dynamics, electromagnetics, structural dynamics, materials science, etc.). Fluid dynamics simulations for realistic aerospace engineering applications often require the computations of complex three-dimensional, unsteady, separated, and turbulent flows. Turbulence modeling plays a vital role in the simulations of these complex flows. Various numerical methods have been proposed in the literature for turbulent flow calculations: direct numerical simulations (DNS), large-eddy simulations (LES), Reynolds-averaged Navier–Stokes simulations, and the hybrid methods, for example, detached eddy simulations (DES), etc., that couple these simulations. As the applications for these simulations expand, the demand for their flexibility and utility grows. Interactive computational steering is one way to increase the utility of these high-performance simulations because they facilitate the process of scientific discovery by allowing the scientists to interact with their data. On yet another front, the rapidly increasing power of computers and hardware rendering systems has motivated the creation of visually rich and perceptually realistic virtual environment (VE) applications. The combination of the two provides one of the most realistic and powerful simulation tools available to the scientific community.

Whereas a tremendous amount of work has gone into developing parallel computational fluid dynamics (CFD) software, little has been done until the past few years to develop parallel computational steering tools that can be integrated with such parallel CFD software. Without such computational steering systems, the immediate and direct interaction with simulations is impossible. Instead, visualization of these simulations is mainly limited to the postprocessing of data. However, for large-scale, parallel aerospace CFD simulations, it is often not possible to postprocess the entire flowfield on a single computer due to memory and speed constraints. Large parallel supercomputers can store thousands of times more data than a typical workstation. Also, the interaction with the large-scale data, which is especially important at the early stages of the computations to tune the several flow and input parameters, becomes very difficult and time consuming for such large-scale simulations. Therefore, interactive parallel (and scalable) computational steering and monitoring systems are necessary to perform efficiently large-scale, parallel aerospace CFD simulations. The system described here allows the steering, visualization, and CFD to all be run in a scalable manner. For example, the majority of the work