Manu Shantharam, SDSC (Original), Steve Lantz (Updates)
Cornell Center for Advanced Computing

Revisions: 11/2022, 12/2020, 7/2017 (original)

PnetCDF is a parallel I/O library developed by Argonne National Laboratory and Northwestern University to manage data stored in netCDF based files (CDF-1, CDF-2 and CDF-5). It provides high-performance parallel I/O capabilities within an application and utilizes MPI-IO as the underlying parallel I/O layer. One should not confuse it with "Parallel NetCDF", which is (strangely enough) built around parallel HDF5.

Objectives

After you complete this topic, you should be able to:

  • Contrast the functionality of netCDF with that of PnetCDF when used in a parallel file system
  • Distinguish between independent and collective calls in the PnetCDF API
  • Create and open files using the PnetCDF API
  • Describe accessing files using PnetCDF
  • Demonstrate querying data using PnetCDF
Prerequisites

To complete this topic, you will need basic knowledge of the netCDF structured data format and basic familiarity with MPI parallel I/O concepts.

 
©   Cornell University  |  Center for Advanced Computing  |  Copyright Statement  |  Inclusivity Statement