TG-ATM130009

NameTG-ATM130009
DescriptionThe aim of this project is to create a global map that describes the functional distributions which characterize the spectraof precipitating auroral particles. Such a map would not only aid in efforts to model the ionosphere and the geospatial environment, but would also aid in the understanding of the magnetospheric source regions of these particles. The construction of this map will utilize programs that perform automated, nonlinear least squares fits of Maxwellian and Lorentzian distributions to data from the Defense Meteorological Satellite Program (DMSP) suite of spacecraft. These programs have been developed in C under the Fedora Coredistribution and are statically linked against the HDF4 and GNU Scientific libraries. Initial testing on an AMD Athlon II X4 645quad core processor has shown that the serial execution of four separate instances of the Maxwellian automated fitting program produces fits at an average rate of 1500 spectra per hour per core, while the Lorentzian program, when executed in the same fashion, will produce fits at an average rate of 500 spectra per hour per core. The 20-year catalog of data which we will use to populate this map contains roughly 200 million spectra. With these average rates, it would take our 16-core cluster approximately 8300 hours to fit Maxwellian distributions and 25,000 hours to fit Lorentzian distributions to the entire catalog. Given this amount of processing time and the serial nature of our programs, we wish to explore the feasability of using your HTC resources to complete this project.
OrganizationUniversity of Texas at Dallas
DepartmentPhysics
Sponsor Campus GridOSG-XSEDE
Principal Investigator
Phillip Anderson
Field Of ScienceAtmospheric Sciences
DisableFalse