Hypernet Labs Co-Founder and CTO Todd Chapman recently had the pleasure of presenting Galileo at the International Conference on Water Management Modeling, which took place virtually on Feb. 24 and 25, 2021. The conference, which showcased 32 presentations and over 80 presenters, provided an excellent venue for the release of the Galileo PCSWMM deployment pipeline.
Built directly into the PCSWMM desktop application, the Galileo deployment pipeline was created by Hypernet Labs in collaboration with Computational Hydraulics International (CHI). Visit the open-source repository here, and try it out with the latest version of the PCSWMM desktop app.
In the presentation, accessible here (video and ppt), Todd explains that containerization and Galileo offer a scaling solution that addresses several major challenges facing engineers working with computationally intensive 2D hydrodynamic models.
These challenges include long run times as well as the need to run many scenarios, to ascertain risk profiles, due to the complexity of boundary conditions. Uncertainty quantification using Monte Carlo analysis, which is likely to become mandatory in the near future, also requires thousands to tens of thousands of realizations for numerical convergence.
“If each run takes multiple days to complete, and you’re trying to do this on a single desktop or maybe two desktops, you’re going to find yourself in a time crunch,” says Todd Chapman.
Todd compares and contrasts the containerization + Galileo solution with several other approaches to accelerating software originally designed for use on local computers. He covers intrusive and non-intrusive approaches, which include model order reduction, distributed memory parallelization, GPU acceleration, and temporary resource scaling.
Containerization is the only method that offers consistency and experimental reproducibility without requiring the rewriting of potentially large amounts of code. However, there’s also a very steep learning curve for the virtualization of new applications (meaning Dockerfiles can be difficult to write), and model management doesn’t come out-of-the-box. Modelers using containers would still need a pipeline for the deployment and monitoring of simulations and retrieval of output files.
Galileo offers the consistency of containerization as well as a ready-made deployment pipeline. In cooperation with CHI, Hypernet Labs has even made the pipeline directly accessible inside the PCSWMM desktop app!
Get it now here, and see it in action at the end of Todd’s presentation video.