Description
In 2020, 35% of users at the National Energy Research Scientific Computing Center (NERSC) used Python on the Cori supercomputer. How do we know? We developed a simple, minimally invasive monitoring framework that leverages standard Python features to capture Python imports and other job data. The data are analyzed with GPU-enabled Python libraries (Dask + cuDF) in a Jupyter notebook, and results are summarized in a Voila dashboard. After detailing our methodology, we provide a high-level tour of some of the data we’ve gathered over the past year. We conclude by outlining future work and potential broader applications.