Ludovic, We don't have a Use Case Diagram for using parallel netCDF, but it's most useful when you need to write into a single netCDF file from a program such as a model running on multiple processors, and more than one processor needs to write to the file at the same time. This occurs a lot in high-performance computing contexts where, for example, a model is running on a large supercomputer and needs to output information at intermediate time steps or for restarting the model without duplicating a lot of work, in case something goes wrong. There are currently two implementations of parallel netCDF, one from an independent project: http://trac.mcs.anl.gov/projects/parallel-netcdf and the other is included in the netCDF-4.0 distribution, implemented over HDF5's parallel I/O: http://www.unidata.ucar.edu/netcdf/docs/netcdf-c.html#parallel-access There is also a section of our online netCDF workshop for developers and data providers that covers both types of parallel I/O: http://www.unidata.ucar.edu/netcdf/workshops/2007/nc4features/Parallel.html http://www.unidata.ucar.edu/netcdf/workshops/2007/pnetcdf/index.html --Russ --Russ Russ Rew UCAR Unidata Program address@hidden http://www.unidata.ucar.edu Ticket Details =================== Ticket ID: NJG-578265 Department: Support netCDF Priority: Normal Status: Closed