ScienceIT has launched a new pilot called Berkelium, Berkeley Lab’s new Kubernetes cluster for building, testing, automating, and sharing scientific work. Researchers can see a demo at LabTech on November 3rd, 2026.
Traditional high-performance computing systems such as Lawrencium remain the heavy hitters for large simulations, parallel computing, and tasks requiring significant computational power. However, modern research increasingly relies on interactive applications, automated workflows, persistent services, and specialized software environments. Berkelium is designed to support this more dynamic, service-driven side of science.
Unlike a traditional HPC cluster, Berkelium is built on Kubernetes, an open-source platform for running and managing applications packaged in containers. A container packages an application together with the libraries, dependencies, and software versions it needs. This allows researchers to use new or older software versions without waiting for systemwide installations or encountering conflicts with other packages.
Because the environment travels with the container, the same experiment can be run on different systems with consistent results. This supports reproducibility, which is an essential part of the scientific process, while giving researchers greater control over their individual computing environments.
Berkelium can support a wide range of projects, including:
- Persistent workspaces: Run JupyterLab or VS Code environments that remain available when you need them, creating a powerful remote workspace close to your data.
- Interactive applications: Transform analyses into web-based tools and dashboards using platforms such as Streamlit or Dash, allowing collaborators to explore results through a browser.
- AI training and model serving: Use GPU resources to train models or create dedicated endpoints that make AI models available to other applications. CBorg currently runs services on Berkelium.
- Automated data pipelines: Move data from laboratory instruments into workflows, databases, and continuous analysis systems in near real time.
- Flexible software environments: Select the exact languages, libraries, and software versions required for a project from extensive repositories of existing container images—or create your own.
Berkelium does not replace Lawrencium. Instead, it complements traditional HPC by providing an adaptable environment for workloads that benefit from interactivity, automation, portability, and persistent services.
The October 15 training will introduce Berkelium, explain how it differs from traditional HPC clusters, and demonstrate how researchers can use containers and Kubernetes resources to support their work.
