To make significant new progress in the understanding and modeling of multi-scale, complex fluid flow phenomena such as turbulence, the entire 4-D structure of the simulated phenomena must be widely, and easily, accessible. To achieve such goals, the proposed project aims to couple database technology with high performance computing. Specifically, we will: (i) Develop a storage infrastructure that provides efficient parallel I/O to archived simulation data at all scales by adapting and improving techniques from recent visualization, database organization and storage architecture research. (ii) Construct a data-driven batch-processing framework that supports many concurrent data-intensive queries. (iii) Develop tools to find "interesting regions" in the data, extract various features, and tag those regions with the relevant metadata. (iv) Develop tools for interactive visualization over the Internet so that data are accessible at display resolution with queries that may request arbitrary sub-regions. Several core databases will serve as content for the proposed research: (i) forced isotropic turbulent flow to address fundamental unsolved questions on the multi-scale and Lagrangian structure of turbulent flow, (ii) a large turbulent channel flow, of interest to engineering fluid dynamics, (iii) simulations of magneto-hydrodynamic (MHD) turbulence of astrophysical relevance, and (iv) a 24-hr daily cycle of the atmospheric boundary layer including day-time buoyant conditions and night-time stably stratified conditions, of relevance to wind energy.
By tightly coupling the database approach to high-performance scientific simulation, the project has the potential to transform the way research in scientific computing of multi-scale phenomena is carried out and employed. If our approach proves successful, it will become routine for state-of-the-art computational datasets to be archived in databases with easy on-line access, flexible user-friendly analysis tools, and efficient coupling with further simulations. We will apply the tools to be developed to study various important fluid flow phenomena, such as turbulent and chaotic motions in neutral fluids, in plasmas, and in the atmosphere.
