Managing job-slot allocation in a multi-VO environment remains a persistent operational challenge for WLCG sites, particularly when each vir- tual organization employs distinct workload-management and scheduling be- haviors. At the RAL Tier-1, more than a dozen VOs compete for heterogeneous resources while relying on different submission patterns, priority models, and pilot-job strategies. Static fair-share targets are therefore insufficient to predict post-disturbance occupancy. We examine the problem from an operator per- spective by combining a summary of the relevant HTCondor controls with a comparison of the refill semantics of CMS, ATLAS, and LHCb. We then use a deliberately small causal toy model, with fixed VO behavior and fixed out- age shape, to isolate the effect of site-controlled correction mechanisms. The results show that faster accounting alone produces only a modest improvement, whereas earlier reclaim, stronger reclaim, and shorter claim persistence mate- rially improve recovery. The largest reduction in fairness gap is obtained by combining these levers while maintaining high utilization.

Associated Links

• Presentation Slides (PDF)
• Indico Agenda