Recurring Meeting of Cleveland Clinic - Stanford University

Date: May 21, 2014

Time: 4:00 PM EST

Means: Conference Call

Attendees:

1. Ahmet Erdemir (Cleveland Clinic)
2. Craig Bennetts (Cleveland Clinic)
3. Joy Ku (Stanford University)

Agenda:

1. Progress on mockup interface development for results retrieval.
2. Test problem to incorporate post-processing.
3. Decide action items for next meeting.
4. Other.

Immediate Action Items:

• Stanford University (Joy & Henry)

  • Implement post-processing script in the workflow to see how it fits to job submission and results retrieval.
  • Identify what to present in WCB 2014, if necessary ask for test problems for demonstration.

• Cleveland Clinic (Ahmet & Craig)

  • Provide links to for updated test problem files, which include post-processing.
  • Provide a copy of cell-scale mesh generation manuscript, which can be used as another test problem.
  • Make scripts Python 2.6.6 compatible to allow simulations on XSEDE.

Notes:

1. Progress on mockup interface development for results retrieval.
   • A prototype results retrieval interface is implemented. The interface is found at the user's own page, specifically at the "My Page" under "My Jobs". The interface provides the simulation jobs under project heading, including job name, progress and links to results files.
   • Ahmet noted that a project specific jobs summary and statistics may be useful at the project site, i.e. similar to download statistics and history. This may also provide a platform to curate a database of abridged solutions.
2. Test problem to incorporate post-processing.
   • Craig added a post-processing Python script in the source code repository. At this moment, everything is hard coded. It is anticipated that the post-processing script will be model specific as in model template file, model configuration file, and model modification script. When a new model is added to cloud computing, all four needs to be provided. At this moment, some intermediate files created during a simulation. Post-processing script reads some of these files and creates an abridged results file for distribution, e.g. post-processed.txt. For results retrieval FEBio's solution and log files, *.xplt and *.log can be compressed as zip and provided as the raw results file and the post-processing results can be provided as the summary results file for the user to quickly evaluate an important result of a job and the simulations appropriateness.
   • The Cleveland Clinic team will provide the links to Stanford University group for the new set of test problem files, which include post-processing.
3. Decide action items for next meeting.
   • See Immediate Action Items above.
4. Other.
   • Henry also setup the job submission interface to connect to XSEDE. XSEDE servers are running Python 2.6; the Cleveland Clinic team will look into their scripts to make them compatible.
   • Joy wants to show a demo of the interface in WCB 2014; e.g. job submission and job retrieval interfaces connected to computing servers. If time permits, an administration site (to provide models for cloud computing) may be developed before the conference. This general framework will be great to show. Joy may need additional problems to illustrate the workflow on a diverse set of models.
   • A concern from the Stanford University group is related to work load to maintain different versions of simulation software packages on compute servers. Currently, this process is semi-automated and relies on Henry as the single maintenance provider. If generalized for a diverse set of projects, this may be overwhelming.

   • The group talked about another test problem, a published study from Cleveland Clinic group to generate cell scale meshes in a scripted fashion. The scripts are available at the relevant project in the SimTk site, https://simtk.org/home/j2c. The scripts require Salome as the unsupervised solid modeling and mesh generation engine, an open source software which can be installed on the compute servers. Cleveland Clinic team will provide the link to the download package and the manuscript that describes the workflow.