Summary (2-4 sentences)
EAF is a JupyterHub-based scientific analysis facility hosted at Fermilab. With more than 35 different environments, it provides access to specialized hardware and cutting-edge machine learning, scientific software and grid utilities for Fermilab experiments, including remote HTCondor submission to our CMS LPC anlaysis cluster. EAF has enabled the use of Dask Gateway, a secure, multi-tenant server for managing distributed python analysis clusters through a hybrid model where task schedulers are executed as Kubernetes pods and task workers are submitted as HTCondor jobs. Once jobs reach the HTCondor worker node, a Dask worker payload is executed inside an apptainer container, which automatically announces itself to the scheduler. This enables EAF analyzers to add large chunks of resources on-the-fly by leveraging the power of HTCondor and Dask for fast, reliable analysis.
Availability of the Speaker
I will NOT be available on July 14th