A Simple Approach to Enhance I/O Performance in the WRF-CMAQ Two-Way Coupled Model

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 7 January 2015
David Wong, EPA, Research Triangle Park, NC; and C. E. Yang, J. Fu, K. Wong, and Y. Gao

I/O is part of a scientific model and it takes up a significant portion of the simulation. There is no exception for the newly developed WRF-CMAQ two-way coupled model at US EPA.  This two-way coupled meteorology and air quality model is composed of the Weather Research and Forecasting (WRF) model and the Community Multiscale Air Quality (CMAQ) model. We are using this two-way model to evaluate how accurate it simulates the effects of aerosol loading on radiative forcing between 1990 and 2010 during which substantial aerosol emissions, such as SO2 and NOx, were reduced in North America and Europe.

The I/O scheme in the current model does not make use of any parallel file system or parallel I/O approach. In addition, the I/O takes about 15% - 28% of the entire simulation. The new approach not only utilizes parallel I/O capability offerred by pnetCDF or netCDF4 but goes one step further to aggregate the data locally, i.e. along column dimension or row dimension among MPI processes in the spatial domain. This approach not only reduces the I/O traffic contention but also aggregated data enhances the I/O efficiency. In terms of I/O time, we have shown this method is about 6 to 10 times faster than the current existing I/O scheme in the model and about 20% - 3 times faster than strict application of pnetCDF. We are currently testing the same idea with netCDF4.