OpenSees from Web Portal#

Hybrid Workflow with Web Portal + JupyterHub

DesignSafe provides a suite of Tapis-powered OpenSees applications accessible through its Web Portal, allowing users to run both sequential and parallel jobs on Stampede3 without writing SLURM scripts manually. This section introduces the available apps, how they work, and how to integrate them into a JupyterHub-driven workflow.

Available Applications#

App Name

Description

OpenSees-Express

Sequential execution on a VM; great for small or test models.

OpenSeesMP

Parallel execution using MPI across multiple nodes for large simulations.

OpenSeesSP

Parallel static analysis (domain decomposition).

These apps are backed by the Tapis platform, which handles job submission, staging, monitoring, and result retrieval from TACC systems.

Why Use the Web Portal?#

The Web Portal simplifies the entire HPC job submission process:

  • Upload input files via a browser

  • Select the app and main input script

  • Configure compute settings (cores, wall time, etc.)

  • Submit without writing a SLURM script

Under the hood, the Web Portal:

  • Generates a SLURM job file

  • Stages input files to Stampede3’s scratch space

  • Launches the job using Tapis + SLURM

  • Transfers output back to your My Data folder

Without Web Portal App

With Web Portal App

Manual SLURM script writing

Script auto-generated

Manual file transfers

Input/output staged automatically

Manual module setup

Preconfigured compute environment

Manual cleanup or downloads

Output copied back to My Data

Note

For short jobs, Web Portal queue times and data transfers may dominate runtime. Use accordingly.

Summary#

The OpenSees Web Portal Apps let you:

  • Submit OpenSees jobs to Stampede3 without learning SLURM

  • Easily switch between sequential and parallel workflows

  • Combine the power of Jupyter for scripting and visualization with the Web Portal for scalable HPC execution

  • Customize or automate workflows via Tapis (for advanced users)

Whether you’re running one job or one thousand, this hybrid approach saves time, reduces errors, and allows you to focus on the science.