This website is a portal to an Open Numerical Turbulence Laboratory that enables access to multi-Terabyte turbulence databases. The data reside on distributed nodes and disks on our database cluster computers and are stored in small 3D sub-cubes to accelerate access speeds to small data subsets.
Access to the data is facilitated by Web Service Interfaces that permit numerical experiments to be run across the Internet. We offer Python and Matlab™ interfaces layered above Web Services so that scientists can use familiar programming tools. Python and Matlab™ codes can be run on client platforms. Python can also be run on Sciserver, a fully integrated cyberinfrastructure system encompassing related tools and services to enable researchers to cope with scientific big data. Subsets of the data can be downloaded in hdf5 file format using the data cutout service. For initial experimentation and familiarization of datasets and access modes, web-browser manual querying at individual points and times is also supported.
Data processing supporting queries for velocity, pressure, gradients, Laplacians, and Hessians (and other relevant variables depending on the dataset) at arbitrary points and time is supported using Lagrange Polynomial and Spline methods executed on the database close to the data. For time-resolved datasets, fluid particle tracking can be performed both forward and backward in time using a second order accurate Runge-Kutta integration scheme.
Available datasets (see datasets description page) include a space-time history of a direct numerical simulation (DNS) of isotropic turbulence (100 Terabytes), a DNS of magneto-hydrodynamic (MHD) turbulence (50 TB), a DNS of forced, fully developed turbulent channel flow (130 TB), a DNS of homogeneous buoyancy driven turbulence (27 TB), a transitional boundary layer flow (105 TB), data from Large Eddy Simulations (LES) of stably stratified atmospheric turbulent boundary layer (20 TB) and (soon to come) two wind farm LES: one in a conventionally neutral flow and another during a diurnal cycle.
Also available are datasets comprising snapshots (spatially but not temporally resolved data) of 40963 DNS of isotropic turbulence (1 snapshot), 81923 DNS of isotropic turbulence (6 snapshots at higher Reynolds number), rotating stratified turbulence (5 snapshots, 5 Terabytes), a high Reynolds number channel flow (11 snapshots, 20 Terabytes) and 20 snapshots of the LES of stably stratified atmospheric turbulent boundary layer.
Recent datasets are being stored using the Zarr storage format with 64-cube “buckets”, instead of the Z-curve ordering used for the legacy datasets. Also, data are being migrated to Ceph-FS storage from the original SQL-database and FileDB storage modalities. New more modular Python and Matlab™ codes use the REST interface instead of the legacy SOAP interface.