High Performance Container Workshop 2025

Containers have been utilized over the years to encapsulate the execution environment from the actual user environment. They are able to benefit legacy infrastructures by leveraging the lower overhead compared to traditional, hypervisor-based virtualization, as well as provide additional flexibility to new HPC and cloud systems . Their portability, reproducibility and distribution capabilities outclass all prior technologies and disrupt former monolithic architectures, due to sub-second life cycles and self-service provisioning.
This workshop will explore the evolving role of Linux Containers in HPC, Cloud, and AI/ML systems.
Key topics will include scalability, performance, and flexible support of workflows, services, and specialized workloads in AI/ML and how AI/ML intersects with HPC. We will review the rapid advancements introduced in AI/ML over the past months and years, examining their implications for HPC.

Additional areas of discussion will include:
- The use of Kubernetes (K8s) adapted to support AI/ML workloads and the potential concerns it raises
- How do AI factories change the dynamics within the EU and how containers are used
- How do NIMs (artifacts to spin up end-to-end training and inference workflows) impact the HPC/AI community and what is to expect?
- The ongoing trend of even more converged clusters supporting diverse workloads.
- Security (memory encryption, microkernel)
- Performance within more and more diverse infrastructure
- Confidential computing and how it impacts the way clusters are build and workloads are modeled.
- Scheduling schedulers within schedulers (Slurm in K8s, Flux)

The workshop will conclude with a call to action for the HPC community: Let’s engage more deeply to ensure HPC continues to thrive alongside AI/ML rather than being overshadowed by it.

---

Organizers:

• Christian Kniep
• Abdulrahman Azab
• Barbara Krašovec
• Andrew Younge
• Eduardo Arango Guiterrez