This is a wide-open question. Obviously, you can look at the codes I mentioned to see how they best laid out their FE data in the HDF5 file. You can also look at the CGNS data layout.
If you are looking into "general" examples of how to write FE data into an HDF5 file then you can look into a couple of projects we've worked on:
(1) https://github.com/PETTT/miniIO
Is a set of I/O kernels for unstructured, structure, AMR, and cartesian meshes; we used these for parallel benchmarks and they use XDMF. These might be more approachable then looking at full production FE I/O libraries.
(2) https://bitbucket.hdfgroup.org/projects/HDFFV/repos/performance/browse/combust_io
Is another FE I/O kernel we plan on using as an HDF5 benchmark.
Additionally, there are best-use practices you can find on the HDF5 webpage. For example, about performance issues,
https://support.hdfgroup.org/HDF5/faq/perfissues.html
and general FAQs about the software,
https://support.hdfgroup.org/HDF5/hdf5-quest.html
would be a good place to start.
Scot
···
On Feb 21, 2018, at 2:11 AM, paul.carrico@free.fr<mailto:paul.carrico@free.fr> wrote:
Hi Scot,
First of all thanks for this feedback and links.
I would like to say that the goal is first to store and work on the data coming from the finite element solver; in my reflexion the post-processing is another stage and I'm looking at xdmf format for that purpose (that can call hdf5 file for "heavy" data).
My questioning is on the structure of the hdf5 file that will contain huge amount of data (mesh and results); in order to be efficient, I'm wondering where I can find "best pratices", articles or usefull information's in the litterature.
Regards
Paul
Le 2018-02-20 16:24, Scot Breitenfeld a Ă©crit :
If you are looking for an FEA "standard" I/O format, it does not exist. CGNS is as close as it gets, but that is for CFD data. There has been some discussion in the past to extend CGNS to include Solid Mechanics, but it never gained traction. For FEA data formats which are geared towards solid mechanics and store the data in HDF5 you have these that I know of:
(1) Exodus (https://gsjaardema.github.io/seacas/)
(2) Silo (https://wci.llnl.gov/simulation/computer-codes/silo)
(3) MOAB (http://sigma.mcs.anl.gov/moab-library/)
(4) XDMF (http://www.xdmf.org/index.php/XDMF_Model_and_Format)
Most vis packages can read these formats and understand that it is FE data.
Scot
On Feb 20, 2018, at 8:36 AM, paul.carrico@free.fr<mailto:paul.carrico@free.fr> wrote:
Dear All,
I'm thinking in using hdf5 format in order to store FEA results prior to additional "in-house" calculations; I'm currently diging into the docs to unterstand how it works, but I would like to know if a standard (or a "best practice") exists in how to make it for that specific topic/field?
The main objective is not to reinvent the wheel 
but to take benefits to years of feedback from the community.
(I'm using a finite element solver for mechanical and thermal analyses, in static/transient condition, and linear/non-linear calculations ... very basic).
Regards
Paul
_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org<mailto:Hdf-forum@lists.hdfgroup.org>
http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5
_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org<mailto:Hdf-forum@lists.hdfgroup.org>
http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
Twitter: https://twitter.com/hdf5
but to take benefits to years of feedback from the community.