Speaker
Description
This contribution investigates carbon aware compute scheduling using the ATLAS experiment at the Large Hadron Collider (LHC) as an example case. The Worldwide LHC Computing Grid (WLCG) consumes approximately 1.25 TWh of energy per year during LHC runs, with computing jobs dispatched immediately regardless of the carbon intensity of the local electricity grid. The Sustainable Queue is presented, a carbon-aware scheduler that delays flexible ATLAS computing jobs to periods of lower carbon intensity using a tiered percentile threshold system. The scheduler requires no machine learning or carbon forecasting and operates entirely on publicly available grid data. Scenario analysis shows that grid decarbonisation improves the queue performance, while increasing CPU capacity alone reduces it. These first results give an idea of constraints and opportunities of temporal shifting in these contexts and are meant to be a stepping stone for more in-depth investigations.