Optimized Utilization of HPC Centers for High-Energy Physics Workflows and Jet Energy Corrections for the CMS Experiment

This thesis presents a comprehensive study that bridges experimental High-Energy Physics with innovative strategies for a more sustainable future of the LHC computing in Germany and beyond.
In the physics domain, the work describes state-of-the-art techniques for precise Jet Energy Calibration for the CMS experiment, with a key contribution being the first full derivation of the absolute residual Jet Energy Corrections for Run 2 Ultra Legacy data.
The developed tools and methods were carefully adapted in close collaboration with the JERC group of the CMS Collaboration to meet the latest reprocessing requirements.
The results were handed over as the final contribution to the Jet Energy Calibration effort from the KIT-CMS group.
Comprehensive analyses demonstrated that the provided corrections minimize the discrepancies between data and simulation to sub-percent levels over large regions of the most relevant phase space, laying ground for high-precision QCD studies and searches for physics beyond the Standard Model -- which both require highest accuracy in jet energy measurements.
In addition, a synchronization process between the well-established MiniAOD-based calibration frameworks employed for Run 2 and the new NanoAOD tools for the Run 3 calibrations presented a full knowledge transfer into the collaboration and ensured consistent results of the next generation calibration workflows.
... mehr

Complementing the experimental achievements, the thesis also tackles the increasing computational challenges inherent in modern High-Energy Physics (HEP) on the brink of the HL-LHC era.
The sheer amount of data and the increasing complexity of the employed analyses and methods in combination with growing requirements on economical and ecological sustainability present major challenges for the future of world's most sophisticated scientific computing grid -- the Worldwide LHC Computing Grid (WLCG).
Starting with an in-depth review of this collaborative distributed computing infrastructure, this thesis provides a comprehensive introduction to the LHC computing, including the dynamic integration of (opportunistic) resources.
Especially the integration of the HoreKa HPC cluster as part of the upcoming HEP computing strategy in Germany, focusing on sustainability and the utilization of HPC resources in the future, is addressed.
By proposing new evaluation metrics, this work lays the groundwork for evaluating and optimizing the grid integration of such resources.
A notable innovation is the XBuffer concept, an XRootd-based approach designed to mitigate bottlenecks in HPC resource utilization.
Despite initial challenges, such as increased failure rates and timeouts, subsequent modifications and optimizations of the concept led to a prototype setup that not only improved reliability and efficiency but also allowed the HPC center to compete with traditional grid sites in reliability and efficiency.
This optimized integration offers a promising outlook for the cooperation between the HEP community and national HPC centers, while also highlighting specific areas for further optimization.

By interweaving precise Jet Energy Calibrations with strategic advancements in scientific computing, this work not only supports the investigation of fundamental interactions at highest accuracy but also paves the way for a more sustainable and efficient computing infrastructure, ultimately bolstering the long-term research capabilities of High-Energy Physics at the LHC.