Interruptible LHCb Monte Carlo jobs

Andrew McNab · CERN Document Server (European Organization for Nuclear Research) · 2018

Overview • Monte Carlo jobs are the bulk of LHCb's offline workload • LHCb has used "elastic" MC jobs for several years • Estimate job time limit and CPU power and calculate how many events will fill the slot • Interruptible MC jobs go further: dynamically stop when time runs out or when external events mean we need to release the slot • Applications in "job masonry" • backfilling the High Level Trigger farm • when sites need to reclaim nodes for kernel updates etc • filling n-processor slots with multiple single processor jobs • And Manchester "eviction" work with Vcycle and/or OpenStack Interruptible LHCb Monte Carlo -Andrew.

Read the paper · More papers on PaperTik