Tools

Software and integrations that help data scientists move data, manage environments, and work with Rivanna and Afton, from your laptop or on the cluster.

Rivanna MCP

UVADS/rivanna-mcp is a local MCP (Model Context Protocol) server that connects AI-enabled IDEs (Cursor, Claude Code, and similar) to Afton/Rivanna over SSH. In plain language, you can ask your assistant about cluster status, your jobs and allocations, storage use, modules, and job submission, without pasting Slurm output by hand.

It is aimed at day-to-day HPC workflow support from your laptop: monitor and manage jobs, inspect resources, and draft or submit batch work through the assistant. Setup, client config, and the full tool list live in the rivanna-mcp README.

Globus file transfer

Globus is RC’s recommended service for moving large datasets to and from HPC storage (/home, /scratch, /project, /standard), lab machines, cloud, and other institutions. Use the managed collection UVA Standard Security Storage in the Globus File Manager; install Globus Connect Personal on a laptop first. For highly sensitive / Ivy data, use the high-security transfer path described in that tutorial and on Ivy/Rio.

Also useful: Globus tutorial · storage notes on RC Services

Apptainer containers

Apptainer (formerly Singularity) runs portable container images on the HPC system. Useful when you need a fixed ML/science stack, Docker-based images, or the same environment in notebooks and batch jobs. RC also ships many apps as container modules. For wiring containers into JupyterLab kernels, see Custom JupyterLab kernels and UVADS/jlab-hpc-containers.

Workflow managers

For multi-step, reproducible pipelines (ETL, bioinformatics, ML training stages, etc.), prefer a workflow manager over a pile of ad-hoc scripts. On UVA HPC these typically submit work through Slurm and can use modules or containers for each step.

Common options for data science and analytics (tutorials and comparisons): UVADS/workflow-basics / site

Tool	Notes
Nextflow	Scalable pipelines; works well with Slurm and Apptainer/Conda on Rivanna/Afton
Snakemake	Make-like Python workflows; strong for file-oriented pipelines
Cromwell / WDL	Workflow Description Language; common in genomics-style pipelines
Prefect	Python-native orchestration for data/analytics workflows

Interactive browser apps (JupyterLab, RStudio, Code Server, Desktop) are covered under Open OnDemand.