Wednesday, August 17, 2022

US-CMS PURSUE internship project: Searching for Extreme Events in Multi-lepton Data from the LHC

 As part of the  US-CMS PURSUE summer internship project, we hosted Xinyue Wu, a rising junior from the University of Rochester. The purpose of the internship is to provide a real research experience to undergrad students in topics related to the CMS particle detector of the Large Hadron Collider. Xinyue implemented a process to search for collision events of interest from real data provided by the CMS detector, and compared the results obtained to predicted yields as computed from simulation data.  The events found were then visualized using tools provided by the CMS software environment.

For this project, Xinyue quickly learned how to use different technologies related to CMS analysis workflows running at scale in a distributed system. These included HTCondor, Coffea, TopCoffea, and Work Queue. Xinyue was able to run this analysis workflow using hundreds of cores at the University of Notre Dame compute nodes.

 We thank Ph.D. candidate Kelci Mohrman and Prof. Kevin Lannon for their support and input. We congratulate Xinyue on the very successful completion of this project!

Link to the final poster

Demo of the Work Queue executor at Coffea User's meeting

Last August 15, 2022 we gave a demonstration on how to use Coffea using the Work Queue executor. Coffea is a framework for pulling together all the typical needs of a high-energy collider physics (HEP) experiment analysis using the scientific python ecosystem. Using Coffea on top of Work Queue, we can automatically manage the resources (such as cores, memory, and disk) to maximize the number of concurrent tasks that the application can run. Further, since Work Queue tasks can be created dynamically once the application is running, we can shape the size of tasks (in this case, the number of collision of particle events) so that they better fit the resources available.

For this demo, we also showcased a new status display for Work Queue when used inside a jupyter notebook: