Investigation of IO Performance in a Weather Model

Details Investigation of IO Performance in a Weather Model Investigation of IO Performance in a Weather Model

Dec 2003

adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2003agufm.u41b0007y&link_type=abstract

American Geophysical Union, Fall Meeting 2003, abstract #U41B-0007

Statistics

Computation

0399 General Or Miscellaneous, 3329 Mesoscale Meteorology, 3337 Numerical Modeling And Data Assimilation, 6309 Decision Making Under Uncertainty

Scientific paper

WRF (Weather and Research Forecasting Model) is a limited-area weather model that has been used intensively for both weather research and prediction. Due to the heavy computational volume, WRF is using multi-layer domain decomposition method to run the model in parallel supercomputing environments. Although there are several performance metrics to measure the overall model behavior; the IO performance has never been thoroughly evaluated. The Hierarchical Data Format (HDF) developed at the National Center for Supercomputing Application (NCSA) at University of Illinois at Urbana-Champaign has become the primary standard file format for storing data from NASA's Earth Observing System (EOS). Since 1999, NCSA has developed a more general and robust data format, called HDF 5, which will support the future demands of Earth Science. HDF5 provides chunking storage of the data, which improves the performance. HDF5 is also easily to hook with external compression packages and make the data storage more efficient and flexible. Furthermore, HDF5 supports the Message Passing Interface (MPI-I/O) standard, which is capable of performing I/O efficiently in parallel computing environments. Currently NCSA is developing an implementation of WRF sequential and parallel I/O modules that reads and writes HDF5 datasets. The module will be available to anyone using WRF as a new option for I/O. In this poster, we report three case studies to compare the performance among NetCDF IO module, sequential HDF5 IO module and parallel HDF5 IO module. In other work, Unidata and NCSA are collaborating to design netCDF4, a new netCDF built on top of HDF5. We hope the current study will provide some insights for scientists, researchers and developers to make decisions in choosing the best file I/O in their computational applications. For more information: HDF Home Page: http://hdf.ncsa.uiuc.edu WRF Home Page: http://wrf-model.org

Affiliated with

Also associated with

No associations

Rating

Investigation of IO Performance in a Weather Model does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.

If you have personal experience with Investigation of IO Performance in a Weather Model, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Investigation of IO Performance in a Weather Model will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFWR-SCP-O-1426477