Welcome and Introduction

• Christian Bolliger (ETH Zurich)
• Maria Grazia Giuffreda (ETH Zurich / CSCS)

Room: HG D 16.2

A Short History of Batch Systems

• Christian Bolliger (ETH Zurich)

Room: HG D 16.2 Batch systems are dating back to the time of punch cards. For most of the computer users are not confronted with them anymore, the processing of tasks is hidden from them. But in the world of high performance computing they play still a crucial role. The history of batch systems is in fact also a history of supercomputing.

Hyper-Converged Cloud Infrastructure at CSCS

• Dino Conciatore (CSCS)

Room: HG D 16.2 This presentation covers the hyper-converged cloud infrastructure at the Swiss National Supercomputing Centre (CSCS), focusing on the integration of Kubernetes (RKE2), ArgoCD, and Rancher for managing and deploying clusters. Rancher handles deployment on bare-metal and HPC nodes, while Harvester manages virtual clusters. ArgoCD automates deployments and ensures consistency, enabling continuous delivery. The combination of Kubernetes, ArgoCD, Rancher, Harvester, and Terraform creates a scalable, adaptable cloud infrastructure. This case study outlines the architecture, workflows, and benefits of this approach.

Leveraging High-Performance Computing for Physics-Based Ground Motion Modelling: Applications in Seismic Hazard

• Maria Koroni (ETH Zurich)

Room: HG D 16.2 This talk presents a framework for ground motion modelling based on physics-based simulations of seismic wave propagation using GPU-based numerical methods. I will highlight efforts to incorporate complex site properties, such as topography and near-surface geotechnical layers, into advanced earthquake simulations. These simulations enable the development of digital twins for Swiss sites to better inform both countrywide and local hazard estimates and further contribute to probabilistic seismic hazard assessment (PSHA).

Managing Slurm on Clusters with Emerging Technologies

• Urban Borštnik (ETH Zurich)

Room: HG D 16.2 User expectations of clusters are changing, meanwhile modern HPC clusters contain resources that are not fully captured by the batch system's cluster model. We present the challenges and developments of bridging the batch system and the cluster to improve the user experience.

Interfacing AiiDA to FirecREST: A RESTful Approach to HPC Workflows

• Ali Khosravi (PSI)

Room: HG D 16.2 In today’s supercomputing landscape, securing both computational resources and results has become increasingly important. Usually, HPC systems are interfaced with users via SSH. However, due to security vulnerabilities and the expensive costs of login servers, many supercomputing centers are now exploring modern alternatives. One promising alternative is to adopt REST API–based solutions to manage HPC resources. While such APIs provide enhanced security and lower operational costs, they also introduce challenges due to increased latency of REST compared to SSH. This performance limitation might presents a challenge for developers of large-scale, high-performance workflow managers, such as AiiDA, to redesign their systems in ways that handles server communication delays efficiently. This presentation will explore how AiiDA, as a high-performance workflow manager, has been adapted to interface with FirecREST—a REST API–based interface for managing HPC resources. I will discuss some of the technical challenges we faced and how we addressed them to maintain performance.

Community Development Room: HG D 16.2

Guided Tours Room: HG D 16.2 We offer two different guided tours: 1. focusTerra - Earthquake Simulator In the permanent exhibition of focusTerra, we explore the causes of earthquakes, watch tectonic plates move throughout our planet’s history and talk about the scientific value of quakes. In the simulator you will experience earthquakes in a safe environment and learn what to do in a case of emergency. 2. Datacentre at ETH Zurich