One HTCondor pool delivers over an ExaFLOP hour cost-effectively for IceCube

The article “A cost effective ExaFLOP hour in the Clouds for IceCube” discusses how an existing HTCondor pool was augmented with resources dynamically acquired from AWS, Azure, Google, Open Science Grid, and on-prem resources to perform science for the IceCube project. The objective, on top of (obviously) the science output, was to demonstrate how much compute can someone integrate during a regular working day, using only the two most cost effective SKUs for each Cloud provider.