Wednesday, June 03, 20093:00 PM - 4:30 PMCNLS Conference Room (TA-3, Bldg 1690)|
A Web-Services Accessible Turbulence Database of Isotropic Turbulence
Professor Charles MeneveauJohns Hopkins University
The importance of high-quality Direct Numerical Simulations of canonical turbulent flows to improve fundamental understanding of turbulence has already been amply demonstrated over the past 3 decades. However, it is also recognized that the developments of tools for high-performance computing now far outstrips developments for analysis of the huge datasets generated. The costly effort to generate large computational datasets is largely wasted if facilities are not also developed to archive the data in a form open to creative experiment and analysis by the whole research community. In this presentation we describe an effort to build a public database system archiving a 1024^4 direct numerical simulation (DNS) data set of the space-time evolution of forced isotropic turbulence. The database contains 1024 frames of velocity and pressure fields in forced, isotropic turbulence, spanning about one large eddy turn-over time scale. The data is obtained by pseudo-spectral simulation in a [0,2π]^3 box with 1024^3 grid points. Dealiasing is done by phase-shifting. The velocity and pressure fields are stored every 10 steps in the simulation. The simulation is performed on a computational cluster. The data is then ingested into the database cluster. A space-filling Morton-curve is used to index the physical space uniformly, and also to organize data partition and distribution. The database system (see htp://turbulence.pha.jhu.edu) allows users access and to process the data remotely through an interface based on the Web-Service model. The users are thus able to perform numerical experiments on the high-resolution direct numerical simulation data using least capable desktop computers. The architecture of the database is explained. Test calculations are performed to illustrate the usage of the system and to verify the correctness of the data. This is a multidisciplinary collaboration involving the groups of the author as well as Profs. R. Burns, A. Szalay, S. Chen and G. Eyink.
Host: Ron Pistone, firstname.lastname@example.org