hpc-ch forum on Improving Access to HPC

Thursday 7 May 2026, 09:30 → 16:50 Europe/Zurich

Conference Center (EPFL - Starling Hotel Lausanne)

Gilles Fourestey (EPFL), Maria Grazia Giuffreda (CSCS)

Overview:

Improving Access to High-Performance Computing (HPC) focuses on making powerful computing resources more available to a wider range of users. Currently, access is limited by high costs, technical complexity, and a lack of skills.

To address this, efforts should focus on expanding shared infrastructure and resource availability, improving education and training so more people can use these systems effectively, and creating supportive policies and funding models to reduce financial barriers. Additionally, making software more user-friendly and encouraging collaboration and resource sharing can further broaden access.

Overall, improving access to HPC means making it easier, more affordable, and more inclusive for everyone to benefit from advanced computing.

Key Questions:

• What are the main barriers (technical, financial, and educational) that limit access to High-Performance Computing for researchers and organizations?
• How can cloud-based HPC services help improve accessibility compared to traditional on-premises supercomputing systems
• How can institutions better prepare users to utilize these resources effectively?
• In what ways can software tools, interfaces, and automation reduce the complexity of using HPC systems for non-expert users?
• How can international collaboration and shared HPC infrastructures (e.g., research networks or consortia) improve equitable access to computing power?

09:30 → 10:00 Welcome Coffee & Registration

10:00 → 10:15 Welcome and Introduction

Conveners: Gilles Fourestey (EPFL), Maria Grazia Giuffreda (CSCS)

10:15 → 11:00 Democratizing HPC: Expanding Access through SCITAS’s Fully Integrated, Data-Centric Infrastructure

Widening access to high-performance computing is essential to unlocking its full impact. At EPFL, SCITAS has adopted a strategic approach to HPC accessibility through a fully integrated, data-centric infrastructure that combines intuitive interfaces with flexible resources. In particular, the introduction of Open OnDemand provides a seamless, web-based entry point to HPC systems, lowering technical barriers for a broad user base. In parallel, SCITAS has extented its ecosystem to the cloud by replicating on-premise environments on demand, enabling users to scale workflows transparently beyond local capacity.

This presentation highlights how usability, environment consistency, and hybrid HPC-cloud integration work together to democratize access to advanced computing and support a more diverse, agile research community.

Convener: Gilles Fourestey (EPFL)

11:00 → 11:30 Democratization of HPC in Action: Empowering Researchers with Natural Language HPC and Quantum Gateways

Building on SCITAS’s integrated infrastructure, this session moves from architectural strategy to real-world application. We demonstrate how the Open OnDemand platform is evolving to meet the needs of the next generation of researchers. By showcasing two novel additions: a seamless gateway to live quantum computing resources, and the "SCITAS Agent," an AI-driven natural language interface, we illustrate a future where technical complexity is abstracted away. This talk highlights how "science by chatting" and push-button quantum access are transforming HPC from a specialized tool into an accessible research ecosystem.

11:30 → 12:00 Bridging HPC and Accessibility: An Open OnDemand Deployment for Diverse Research Communities

Open OnDemand (OOD) is a web-based platform that democratizes access to high-performance computing resources through any browser-enabled device.

At the University of Lausanne (UNIL), we deployed and extensively customized an OOD instance to serve a diverse research community with varying levels of technical expertise. Our deployment addresses the challenges of a highly heterogeneous cluster environment, integrating SLURM scheduling and complex software stacks managed through Spack and Apptainer/Singularity containerization.

We present our deployment strategy, including integration of custom tooling for large-scale data movements between on-premise and cloud storage systems, as well as interactive web applications targeting scientific computing and teaching (Jupyter, RStudio, containerized desktop environments).

We conclude with community feedback and adoption metrics that highlight both successes and ongoing challenges in supporting non-technical users on shared HPC infrastructure.

Convener: Flavio Calvo (University of Lausanne)

12:00 → 12:30 HPC and User Behaviour: The Customer is Always Right

HPC admins often encounter users misusing resources, running too many processes in the login nodes, avoiding reading the documentation and being unprepared to work in an HPC cluster. However, what if they are just a reflection of our setup? At PSI, we try to understand our user's actions and backgrounds to create and adapt tools to shape their behaviour by providing lower friction choices and incentives for demanding tasks.

Speaker: João Agostinho de Sousa (PSI)

12:30 → 13:30 Lunch and Networking

13:30 → 14:00 Lowering the Barriers to HPC

Traditional HPC systems often have technical or administrative barriers to use, ranging from account requests to using software. We describe how we strive to lower unnecessary hurdle but also highlight the challenges arising from lower barriers to entry.

Convener: Urban Borštnik (ETH Zurich)

14:00 → 14:30 Bridging Cloud and HPC Workloads with the Renku Platform

Data science projects combine code and data from diverse locations, and compute platforms such as cloud and HPC. Coordinating these resources requires technical knowledge and represents time-consuming overhead, diverts time from core analytical work and impedes collaboration. When resources are not brought together in a repeatable, structured manner, team members struggle to replicate software environments or access necessary datasets, ultimately slowing project progress and hindering knowledge sharing.

Renku addresses these challenges by providing an integrated platform that connects data, code, compute in a unified workspace allowing for seamless transitions between commodity cloud or specialized HPC resources from the same easy-to-use web interface. The platform enables data scientists to launch browser-based interactive computing sessions with software dependencies pre-configured and data and code automatically accessible, thus eliminating setup friction and facilitating seamless collaboration.

Built by the Swiss Data Science Center, Renku is an open source collaboration platform used nationwide for data science projects, research, and teaching. By streamlining the technical infrastructure of data science projects, Renku allows teams to focus on analytical insights rather than resource management, empowering collaborative knowledge building.

Conveners: Rok Roškar (SDSC), Salim Kayal (Idiap)

14:30 → 15:00 HPC Console - a Lightweight Web-UI for HPC and AI

Command-line access to HPC systems is powerful but increasingly a barrier for today's diverse user base. This talk introduces an open-source, web-based HPC Console that offers a unified interface for system monitoring, job submission and management, and data handling across heterogeneous environments.
Built on top of the FirecREST APIs, the Console acts as a secure intermediary between users and HPC resources — providing scheduler-agnostic access to compute, storage, and monitoring services without exposing credentials or requiring direct shell interaction. Its architecture emphasizes server-side rendering, route-centric data orchestration, and strong separation of concerns for consistency, security, and scalability.

We'll walk through key features including cluster health awareness to prevent unsafe job execution, multi-cluster and multi-account support, contextual job monitoring with pluggable observability integrations, and streamlined workflows for moving large datasets.

Convener: Elia Palme (CSCS)

15:00 → 16:00 Guided Tours

• Tour I: CCT-DC tour (Centrale de chauffe thermique)
  Discover EPFL's cutting-edge CCT-Data Center, a 1,000m² facility hosting 200 water-cooled racks and powered by renewable energy from photovoltaic panels and an innovative heating system using water from Lake Geneva. Currently home to supercomputers JED and KUMA, the center supports advanced scientific research, particularly in Artificial Intelligence and deep learning. KUMA stands out as an environmentally efficient resource, ranking 23rd in the Green500 and 103rd in the Top500 most powerful supercomputers in 2024, showcasing EPFL's commitment to innovation and sustainability.

• Tour II: SPC/TCV Tokamak
  Explore the Swiss Plasma Center (SPC) and dive into the world of plasma and fusion research during this exclusive tour. Discover cutting-edge facilities, including Switzerland’s only tokamak, a state-of-the-art fusion device using magnetic fields to confine plasma within a toroidal chamber. Learn about groundbreaking advancements in fusion energy at one of the world’s premier research centers.

CCT-DC_external larger.jpg

Tokamak.png

16:00 → 16:05 Farewell and End of the Meeting