Parallel-NetCDF included in NetCDF-4 ?

[Ed Hartnett]

Let me summarize some HPC options for I/O.

pnetcdf
********

A stand-alone, netcdf-compatible C/Fortran library from Argonne.

PROs:
* good performance
* good access to MPI features.
* some extra buffering for data which is helpful for HPC.
* fastest/simplist way to get decent HPC performance, and netcdf output.

CONs:
* slightly different API from netcdf,  so code is not-quite compatible with
standard netcdf code.
* classic formats only (including new CDF5 format, which allows very large
dimensions), but no netcdf/HDF5 formats.

Netcdf/HDF5
**************

The standard netcdf library, built on a parallel system (with mpicc)
(C/Fortran/C++):

PROs:

* standard API, code compatible with all standard netcdf code.

CONs:

* does not provide pnetcdf-style extra buffering.
* only netcdf/HDF5 output format supports parallel IO, unless built with
--enable-pnetcdf.

NCAR PIO library
******************

A C/Fortran library which uses the netcdf library and (optionally) pnetcdf,
to provide additional HPC features :

PROs:

* supports pnetcdf performance.
* allows a subset or a disjoint set of processors to be used for IO.
* supports classic and netcdf/HDF5 formats.
* supports distributed arrays.

CONs:

* slightly different API from netcdf standard.
* extra set-up for distributed arrays.

Keep on NetCDFing!

Ed Hartnett

On Jun 7, 2017 1:42 AM, "Wei-keng Liao" <wkliao at eecs.northwestern.edu>
wrote:

> Hi, Louis
>
> You can use either NetCDF or PnetCDF library to write/read files in
> parallel.
> Please note the APIs are different between the two libraries.
>
> If the NetCDF library is built with PnetCDF support, then you can write a
> program using NetCDF APIs to perform parallel I/O. In this case, the
> parallel
> I/O is carried out by PnetCDF library underneath.
>
> Similarly, the NetCDF library can be built with HDF5 support and the
> parallel
> I/O is carried out by HDF5 library underneath.
>
> The software release of NetCDF libraries do not include PnetCDF source
> codes
> nor HDF5 source codes. You must obtain and build them separately.
>
> There are two kinds of file formats supported by NetCDF library, the netCDF
> classical file format and HDF5 format. When using PnetCDF underneath, the
> file format will be classical and using HDF5 underneath, the HDF5 format.
>
> There are some parallel I/O features in PnetCDF that have not appeared in
> NetCDF yet, such as request aggregation through nonblocking APIs. The
> aggregation feature often result in a good I/O performance.
>
> Let me know if you have further questions.
>
> Wei-keng
>
> On Jun 6, 2017, at 1:49 PM, Monier, Louis wrote:
>
> > Dear Parallel-NetCDF users,
> >
> > I’m Louis Monier, a young undergraduate from France who joined Zhiming
> Kuang’s research group in Harvard university for 4 months (internship). I’m
> working on the optimization of a  code (related to atmospheric modelling)
> with the use of parallel NetCDF to speed up the output process.
> >
> > I’m a beginner in parallel computing. I would have a important question
> for you. I can’t find a clear answer on the internet:
> >
> > I’ve seen on the Parallel-NetCDF web site (http://trac.mcs.anl.gov/
> projects/parallel-netcdf) that starting from version 4.4.0, the NetCDF
> library includes both sequential and parallel access. So here is my
> question:
> > If I want to speed up my output process using a NetCDF file with a
> parallel write process, should I use the PNETCDF library or NETCDF library
> (starting from version 4.4.0) ?
> > In other words, is PNETCDF library included in NETCDF library from
> version 4.4.0 ?
> >
> > Indeed, it’s possible to find some parallel-NetCDF examples. Some are
> using NETCDF library, other are using PNETCDF library. I don’t which one is
> best to use.
> >
> > Thank you to inform me on this subject.
> >
> > Best,
> >
> > Louis Monier
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/parallel-netcdf/attachments/20170607/038c570b/attachment.html>