Vine CMS Analysis Stack: Tutorial & Hackathon

Created on November 02, 2026 by Cooperative Computing Lab

2026   ·   tutorial   cms   ccl   lab-updates   taskvine   ·   news

We are excited to share the Vine CMS Analysis Stack, a platform designed to accelerate your computations and streamline your research workflow (and help graduate students finish sooner!).

Built on years of experience scaling data analysis software for the CMS experiment, this open-source stack has enabled Notre Dame researchers to execute analysis workloads across thousands of cluster nodes simultaneously. We are now expanding our reach to deploy this framework with CMS collaborators worldwide.

On November 12th and 13th, the Cooperative Computing Lab (CCL) is hosting a hands-on Tutorial & Hackathon tailored for Notre Dame CMS graduate students, postdocs, and faculty.

Schedule

  • Thursday, November 12, 10:00 AM–1:00 PM: Tutorial
  • Friday, November 13, 2:00–5:00 PM: Hackathon

Location: Nieuwland Science Hall, University of Notre Dame (room TBA). Remote participation will also be supported.

Who Should Attend

Anyone comfortable with a terminal, Python, and a text editor. No prior experience with TaskVine, coffea, or HTCondor is needed.

Tutorial (Thursday)

You’ll take a real CMS analysis from a single process on a login node, to a TaskVine worker on the same machine, and finally to a full HTCondor pool, all without changing your processor code. Along the way, you’ll learn how to:

  • Run the same analysis locally, on a local TaskVine worker, and on an HTCondor pool
  • Package a conda environment so remote workers can run your code
  • Read a run’s health from the live console and the on-disk logs
  • Tune chunk sizes, reduction sizes, and resource declarations
  • Use checkpoints to restart a run without redoing finished work
  • Keep a run going when some chunks fail
  • Run several processors over the same datasets in one job

The analysis is written in both coffea and ROOT’s RDataFrame, and you pick whichever you prefer. Both are installed in the same environment, so you can switch at any time.

All exercises use real CMS Open Data from Run2012B_SingleMu, the input to the IRIS-HEP ADL benchmarks: 15 ROOT files and about 3.7 million events.

Hackathon (Friday)

Bring your own analysis code! Adapt and scale your workflows with direct support from the CCL team.

Before the Session

Please complete these steps ahead of time. They won’t be covered live, and building the environment can take up to 30 minutes.

  • Have a glados account.
  • Have conda or pixi working.
  • Clone the repository and build its environment by following Module 1: Setup.
  • Pick an analysis flavor: coffea or RDataFrame.

Resources & Registration