Background
I received a BS and BA in Geology and Physics from Rutgers University in 1996. I then attended the University of Texas at Austin and received an MS in the Geosciences in 1998. My thesis (in close collaboration with The Univ. of Texas Institute for Geophysics) involved a geologic and tectonic interpretation using shipborne magnetic data as well as a data processing technique for removing cultural noise from aero-magnetic data collected over the North Slope of Alaska.

After my MS, I worked for Fugro Airborne Surveys (Formally World Geoscience) as a Geophysicist and then for Fugro McClelland Marine Geosciences as a Unix System's Administrator. This was the start of what would be a shift in my career from strictly geology and geophysics to a combination of the geosciences, parallel processing and high performance computing. In the fall of 2001 I began my studies of parallel processing and high-performance computing at the University of Colorado, Boulder, received my MSEE in 2004 and defended my PhD in March 2007.

Research Abstract
My research presents the data relationships necessary to discover and implement a model based load index (MBLI) for load balancing scientific applications on distributed parallel systems. An MBLI is an alternative quantity to run-time measurement-based load indices (RLIs) such as processing time. This newly characterized index must be a quantity produced by or required of the scientific model being simulated.

An MBLI correlates with a measured process performance parameter that directly represents heterogeneous computational loads and can be used to resolve load imbalances that reduce an application's time to completion. The method of obtaining an MBLI occurs during a pre-processing step and does not incur a run-time cost after implementation. Atomic mass, temperature tendency and surface flux are examples of MBLIs found in Molecular Dynamics (MD) models, Atmospheric General Circulation Models (AGCM) and Ocean Circulation Models (OCM) respectively.

This research presents the discovery processes for MBLIs in AGCMs, MD models and OCMs. MBLI implementations and performance of an AGCM and MD model (NAMD2) are discussed while executing on Pentium4 Xeon, IBM Power5-p575 and IBM BlueGene/L systems. The AGCM implementation includes results from both a production model, the Community Climate System Model/Community Atmosphere Model (CCSM/CAM3), and from a Load Balancing and Scheduling Framework (LBSF). A particular LBSF implementation includes the first use of Very Fast Simulated Annealing to make load balancing decisions using an MBLI. Finally, a detailed analysis is presented that compares an RLI to an MBLI and clearly shows the overhead and error associated with a run-time measured quantity.

This work is being carried out under the supervision of Dr. Daniel Connors (Univ. Colorado Boulder), Dr. James J. Hack(NCAR (National Center for Atmospheric Research),CGD/CMS) and Dr. Gita Alaghband (Univ. of Colorado, Denver).