Wednesday, October 12, 2022

Mufasa: Robust Meta-Workflow Management at eScience 2022

This week, I'll be presenting Ben Lyon's M.S. project work on Mufasa at the eScience 2022 conference in Salt Lake City.   Mufasa is a meta-workflow manager that schedules multiple workflows onto a cluster.  When executing a large number of workflows, the WMS itself can have substantial resource consumption on the head node, network, and shared filesystem as it moves data to and from the cluster.  Mufasa schedules these resources, ensuring that the system maintains throughput even under heavy load of arriving workflows.

Friday, September 16, 2022

pyHEP 2022: Automatic resource management with Coffea and WorkQueue

pyHEP is a virtual workshop to discuss the use of python in the High Energy Physics community. In this year pyHEP (2022), we showcased the resource management features that Work Queue offers when executing HEP analysis workflows using the Coffea framework. We showed how Work Queue is able to automatically tailor the resources allocated (cores, memory, disk) to the tasks executed, and how it modify the size of the tasks so that they can fit the resources available.

You can access the jupyter notebook we used for the talk here: