92nd American Meteorological Society Annual Meeting (January 22-26, 2012)

Monday, 23 January 2012: 2:15 PM
Examples of Mosaic Data Processing in Parallel Using CDAT
Room 346/347 (New Orleans Convention Center )
Alexander Pletzer, Tech-X Corp., Boulder, CO; and D. Kindig, S. Vadlamani, C. Doutriaux, J. Painter, and D. Williams

Poster PDF (3.9 MB)

Recently, the Climate Forecasting (CF) conventions have been extended to support GRIDSPEC data spread over multiple files, covering both time and space aggregation. These extensions cover mosaic data on the cubed-sphere, in particular, with its six tiles connected in unstructured fashion. Mosaic data are ideally suited for parallel processing, either on a per tile basis or within a tile using a standard domain decomposition. We show examples of parallel processing using the Python Climate Data Analysis Tools (CDAT) software for a use-case involving interpolation over a region containing a corner point between three tiles. The examples leverage CDAT's recently added capability to understand GRIDSPEC data layouts, and the ability of CDAT variables to interact with C code (LibCF) and play with the Python mpi4py module.

Supplementary URL: