New simulation software technologies at the LHCb Experiment at CERN

TL;DR

The paper addresses the challenge of generating large-scale Monte Carlo simulations for LHCb Run 3 within constrained computing resources due to a fivefold increase in luminosity. It introduces Gaussino as an experiment-independent core that unifies Gaudi with Geant4 in a multi-threaded framework, and Gauss-on-Gaussino as the LHCb-specific integration that exposes a FastSimulation interface to support fast and ultra-fast models, including DL-based approaches. Training-data pipelines using ExternalDetector and simple abstract models (ImmediateDeposit, ShowerDeposit) are described to enable practical development of fast simulations for upgrade geometries. The work demonstrates substantial potential CPU-time reductions for fast paths while preserving data integrity, laying the groundwork for scalable, DL-enabled fast simulations in the LHCb software stack.

Abstract

The LHCb experiment at the Large Hadron Collider (LHC) at CERN has successfully performed a large number of physics measurements during Runs 1 and 2 of the LHC. Monte Carlo simulation is key to the interpretation of these and future measurements. The LHCb experiment is currently undergoing a major detector upgrade for Run 3 of the LHC to process events with five times higher luminosity. New simulation software technologies have to be introduced to produce simulated data samples of sufficient size within the computing resources allocated for the next few years. Therefore, the LHCb collaboration is currently preparing an upgraded version of its Gauss simulation framework. The new version provides the LHCb specific functionality while its generic simulation infrastructure has been encapsulated in an experiment independent framework, Gaussino. The latter combines the Gaudi core software framework and the Geant4 simulation toolkit and fully exploits their multi-threading capabilities. A prototype of a fast simulation interface to the simulation toolkit is being developed as the latest addition to Gaussino to provide an extensive palette of fast simulation models, including new deep learning-based options.

Figures (11)

• Figure 1: Simplified view on the LHCb data flow and data processing applications as used in Run 1 and Run 2.
• Figure 2: Dependencies FastSimInterface in the simulation software stack before and after upgrade.
• Figure 3: Projection of the computing resources available to LHCb LHCB-FIGURE-2019-018. Monte Carlo simulations will quickly exceed in the upcoming years the resources allocated to LHCb by the distributed computing system provided that no action is taken in order to fit in the requirements (e.g. fast simulations). The CPU needs for producing 100% of the required samples with the detailed simulation is shown by the green bars. The CPU needs in two scenarios with different fractions of events produced with detailed, fast and ultra-fast simulations are shown by the blue and dashed purple lines respectively.
• Figure 4: Relative time FastSimInterface spent in each sub-detector when simulating 100 minimum bias events using different versions of the framework with the beam conditions as foreseen in the Run 3 data-taking period and the upgrade geometry. The time spent in the calorimeters is very similar in the current and new versions of the framework as expected. Simulation of the optical photons in RICH1 and RICH2 was turned off in the current version of Gauss as it is not yet available in Gauss-on-Gaussino, for easier comparison.
• Figure 5: A simplified model FastSimInterface of the FastSimulation interface with a set of dedicated factories that construct the corresponding Geant4 objects.