FEniCS

1. Demo MSO4SC Unicorn/FEniCS-HPC Portal workflow with source code web editing

1.1. Contact:

  • Johan Jansson (jjan@kth.se)

  • Cem Degirmenci

  • Javier Carnero (Portal)

1.2. Demo workflow

We have successfully tested the following open source Unicorn/FEniCS-HPC workflow in the MSO4SC Portal with source code web editing in the GutHub editor. The workflow is specifically of critical importance for the online course MOOC-HPFEM, and MOOCs in general. By "user" we mean Portal user and include both someone who simply wants to edit basic information or data and run simulations, and also a developer who wants to carry out extensive source code development:

  1. Branch repository The user branches or forks the Unicorn/FEniCS-HPC Git repository bitbucket.org/fenics-hpc/unicorn , "next" branch.

  2. Source code web editing The user edits the source code and/or data of her own repository in the integrated GitHub web editor: GitHub repository source code web editing.

  3. Run experiment in Portal on FTII The user runs her own repository version of the experiment in Portal-Experiments on FTII, here the wing benchmark, with logging. The repository location is specified in the Experiment form: Run repository version of wing benchmark in Portal-Experiments.

  4. Visualize in Portal The user explores the solution in ParaView the remote desktop Portal-Visualization: Portal-Visualization with remote desktop.

  5. Repeat Based on the conclusions from the visualization, the user may want to edit again and restart the workflow.

1.3. Proposed improvements:

The following proposed improvements should be easily implementable in the current framework, and should elevate the user experience significantly:

  1. GitLab has a more capable web editor.

  2. The web source code editing and repository forking/branch can be more integrated in the Portal user interface, for example automatically forking and opening the web editor, as discussed in Issue: github.com/MSO4SC/MSOPortal/issues/110

  3. Notification of start/end of the job by email.

  4. A default visualization can be done by ParaView in batch mode, and emailed to the user.

  5. Visualization data files can be made available to a data repository (e.g. Google Drive) and visualized in the Portal using ParaViewWeb/ParaView Glance.

  6. Automated MSO4SC run of test experiment on repository commit with the CI/CD framework in MSO4SC.