MONC, the Met Office/NERC Cloud simulator, is a large-eddy simulation code for cloud and atmospheric modelling. MONC is a rewrite of the 1980s/90s Met Office Large Eddy Model and, like many physical and environmental simulation codes, it models the world (the atmosphere in this case) on a large three-dimensional grid. For MONC, “large” means billions of grid points, and billions of grid points means billions of data variables which change over time as the simulation proceeds. This is too many to write out to disk for analysis and visualisation: it would simply take too long. Working with the Met Office, EPCC has developed in situ data analytics capability in MONC, where around one processing core in every nine is dedicated purely to real-time analytics. Neighbouring cores send their simulation data to the analytics core and simply carry on computing. Disk i/o is minimised, and MONC now runs efficiently across over 100,000 cores.