Erlangen 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 14: Data, AI, Computing, Electronics II
T 14.6: Talk
Monday, March 16, 2026, 17:30–17:45, KH 02.014
Background Jobs for Efficient Resource Utilisation on HPC Systems — Inga Lakomiec1, •Ughur Mammadzada1, Saidev Polisetty1, Arnulf Quadt1, Rodney Walker2, and Sebastian Wozniewski1 — 1Georg August University of Göttingen — 2Ludwig Maximilian University of Munich
The upcoming High Luminosity Large Hadron Collider (HL-LHC) era will greatly increase the computational and storage demands of high energy physics beyond what the current Worldwide LHC Computing Grid (WLCG) can provide. In Germany, university based Tier-2 sites are therefore transitioning their workloads to national HPC facilities at Nationales Hochleistungsrechnen (NHR). In this context, the GoeGrid Tier-2 site at the University of Göttingen is migrating its computing activities to the local NHR cluster EMMY.
To integrate HPC resources into the WLCG, virtual worker nodes are created. They must stop accepting WLCG jobs well before the time limit set by the HPC system, which prevents premature termination but leaves a significant fraction of the allocated CPU time unused near the end of the lifecycle. As the HL-LHC will increase the data volume by a factor of ten, such resource loss becomes increasingly inefficient. The heterogeneous WLCG-HPC environment also requires robust monitoring.
This work investigates opportunistic background jobs at GoeGrid and EMMY to recover this otherwise wasted capacity, and reviews the monitoring infrastructure and the concept of virtual worker nodes.
Keywords: HL-LHC; HPC; WLCG; NHR; ATLAS