Solutions

CLIP offers three services. Batch and interactive computing both run on CBE, the CLIP Batch Environment; grid computing serves international particle physics collaborations.

Which option should I use?

• Batch computing: your work runs as scripts without supervision, repeats over many inputs, or runs for a long time.
• Interactive computing: you want to explore data, develop code, or use a graphical application in your browser while the session runs on cluster resources.
• Grid computing: you take part in the ALICE, Belle II, or CMS collaborations, which process data through the grid.

Comparing CLIP with your workstation or with MUSICA? See where CLIP fits.

Batch computing

CBE gives you terminal access to a cluster with CPU nodes, GPU nodes with NVIDIA B200 and A100 GPUs, and high-memory nodes; see Data Center for details. SLURM, an open source job scheduler, places your jobs on compute nodes as resources become available. CBE supports high-throughput computing (HTC): it handles large volumes of computational tasks over extended periods, rather than solving a single problem as fast as possible.

Key features:

• Secure Shell (SSH) access with short-lived certificates instead of passwords. Each certificate is valid for 16 hours.
• A shared scientific software stack from EESSI, the European Environment for Scientific Software Installations, which you load as environment modules.
• GPU jobs that request a full GPU or a smaller, isolated slice of one through Multi-Instance GPU (MIG).
• Job arrays for running the same analysis over many inputs.

Typical workloads include Nextflow pipelines, AlphaFold protein structure prediction, R and Bioconductor genomics analyses, TensorFlow, and the Python scientific stack.

Interactive computing

OpenOnDemand is CBE’s web portal for interactive work. From your browser, you can open a shell, manage files, and start applications such as Jupyter notebooks, RStudio, and remote desktops. Each session runs as a SLURM job on cluster resources, so you choose the runtime, CPU, memory, and accelerators it needs. OpenOnDemand replaces the separate JupyterHub, RStudio Server, and XPRA services of the previous system.

Grid computing

CLIP operates T2_AT_VIENNA, Austria’s only Tier 2 grid computing site, processing jobs for international particle physics collaborations: ALICE, Belle II, and CMS. If you work in one of these collaborations and have a question about the site, email contact@clip.science.

Common tasks

The CBE documentation covers each task in detail:

• Run your first job
• Connect with SSH
• Submit and manage SLURM jobs
• Submit GPU jobs
• Find and load software with EESSI
• Choose storage and transfer data
• Use OpenOnDemand web applications
• Troubleshoot problems and get support

Get help

Account holders can open a ticket in the HPC Servicedesk. Prospective users can email contact@clip.science. See Support for all options.