Efficient Parallelization of a Risk Management Model on the NT Supercluster
B. S. Minsker,* D. E. Goldberg, E. Moosbrugger
Sponsor: University of
Illinois
Methods for efficiently implementing a risk management model on a distributed cluster of commodity computers are being investigated. The model combines a noisy genetic algorithm, which searches for cost-effective corrective action plans, with a flow and transport model called RT3D and a human health exposure risk assessment module. Commodity computers were chosen because the ultimate users of the model will be practitioners and government regulators, who may not have access to massively parallel supercomputers. The research will investigate both single- and multiple-population approaches to genetic algorithm parallelization, exploring innovative methods that should simultaneously improve computational efficiency of the model.
| Home
| Research Areas
| Publications
| Group Members
| Education
| Calendar |
