Abstract | ||
---|---|---|
This poster presents an overview of and the performance results of recent I/O advancements in the parallel CMAQ framework. These optimizations were developed as part of a collaboration between the EPA and Sandia National Laboratories. netCDF provides a portable file format and an easily understood API, but it does not support concurrent writes by multiple processes. In a cluster environment, this leaves two basic options: create a file per process or funnel all the data through a single node. Neither of these options are optimal. Using a slightly modified API, parallel-netCDF (pnetCDF) enables high performance parallel I/O using the MPI-IO collective I/O optimizations while maintaining the netCDF file format. We have created a thin wrapper around pnetCDF that makes it simple for users to enable the new parallel I/O features at compile time. Using parallel I/O has improved the write performance of the CMAQ air-quality modeling code by up to 48%. |
Year | DOI | Venue |
---|---|---|
2006 | 10.1145/1188455.1188631 | SC |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Todd Kordenbrock | 1 | 60 | 4.11 |
Ron Oldfield | 2 | 408 | 18.71 |
Jeffrey Young | 3 | 7 | 2.07 |