Multicore workflow characterisation methodology for payloads running in the ALICE Grid

Abstract

For LHC Run3 the ALICE experiment software stack has been completely refactored, incorporating support for multicore job execution. Whereas in both LHC Run 1 and 2 the Grid jobs were single-process and made use of a single CPU core, the new multicore jobs spawn multiple processes and threads within the payload. Some of these multicore jobs deploy a high amount of shortlived processes, in the order of more than a dozen per second. The overhead of starting so many processes impacts the overall CPU utilization of the payloads, in particular its System component. Furthermore, the short-lived processes were not correctly accounted for by the monitoring system of the experiment. This paper presents the developed new methodology for supervising the payload execution. We also present a black box analysis of the new multicore experiment software framework tracing the used resources and system function calls issued by MonteCarlo simulation jobs. Multiple sources of overhead in the lifecycle of processes and threads have thus been identified. This paper describes how the source of each was traced and what solutions were implemented to address them. These improvements have impacted the resource consumption and the overall turnaround time of these payloads with a notable 35% reduction in execution time for a reference production job. We also introduce how this methodology will be used to further improve the efficiency of our experiment software and what other optimization venues are currently under research.