The TH–NGT Hackathon: Exploring the NGT Cluster

From 23-26 September, TH–NGT hackathon brought together members of CERN’s Theoretical Physics department and the Next Generation Triggers (NGT) project for a hands-on introduction to the NGT Cluster. The event, held at the CMS Centre, aimed to familiarise participants with the cluster’s infrastructure and support the first steps in developing workflows on this new computing resource.

The hackathon opened with an overview of the cluster’s managing system and environment, before shifting focus to NGT work packages WP1.4 and WP1.5, both of which address workflow development and software implementation. By working side by side, participants could test ideas, share knowledge, and progress on technical tasks in a collaborative setting.

The NGT Cluster, developed by Task 1.1 and installed at the Prévessin data centre, is a collection of shared computing resources designed to meet the demands of the High-Luminosity LHC (HL-LHC). It combines a variety of hardware—including 188 GPUs, 4,600 CPUs, and FPGAs—and supports a broad range of interfaces tailored to different workflows. Users can interact with the system through Jupyter notebooks, SSH, or distributed job submissions, enabling both small-scale testing and large-scale machine learning training for experiments like ATLAS and CMS.

During the hackathon, most of the groups worked on projects related to lattice QCD and tensor networks, two important approaches in theoretical physics that require significant computing power.

Lattice QCD is a method used to study the strong interaction—the fundamental force that binds quarks and gluons into protons, neutrons, and other hadrons—by placing them on a discrete space-time grid. This numerical approach allows physicists to calculate properties of strongly interacting matter that are otherwise impossible to solve analytically, but it demands massive computational resources.

Tensor networks, on the other hand, are mathematical structures used to efficiently represent and simulate complex quantum systems. By organising information into interconnected tensors, they make it possible to capture correlations in large systems that would otherwise be computationally intractable. They are increasingly applied in high-energy physics to explore quantum field theories and connections to quantum computing.

Hear from the participants what usage they gave to the Cluster:

By combining technical training with practical experimentation, the hackathon strengthened connections between TH and NGT while preparing participants to make effective use of the cluster. As development continues, the NGT Cluster will play a key role in supporting machine learning and other advanced computing applications essential for HL-LHC physics.