Facilities

The computing we build, operate and keep running: ATLAS production at Tier 2 scale, interactive analysis for physicists, the platforms and services behind them, and a public home for open data.

Midwest Tier 2 Center

ATLAS production computing on the Worldwide LHC Computing Grid

A U.S. ATLAS Tier 2 center since 2005, run as one facility across three universities — Chicago, Indiana, and Illinois inside NCSA's National Petascale Computing Facility — delivering simulation, reconstruction and analysis for the ATLAS experiment at CERN.

  • 42,000 cores About 21,000 physical, roughly 646,000 HS06, scheduled with HTCondor
  • 26 PB on dCache Around 88 million files under management, served at Chicago and Indiana
  • 9.2 million jobs a quarter At about 93% success, moving roughly 40 PB over the research network
Read more →

UChicago Analysis Facility

Interactive analysis at scale for U.S. ATLAS

In production since 2021, the Analysis Facility gives ATLAS physicists login nodes, notebooks, batch computing, GPUs and petabytes of storage, built on Kubernetes and designed around how people actually work.

  • 141 worker nodes 10,444 CPU cores and 42.7 TB of memory in one Kubernetes cluster
  • 8,600 batch cores With about 13,000 more reachable by flocking to MWT2
  • 5.2 PB Ceph-backed user storage, about 1.7 PB usable at three-way replication
Read more →

Hinds Data Center

The R&D platform, the collaboration clusters and the services layer

Three Kubernetes clusters in the Hinds building, where new analysis patterns are proven, where Physical Sciences Division and neutrino physicists compute, and where the identity, registry and data management services other facilities depend on actually run.

  • River The IRIS-HEP Scalable Systems Laboratory testbed, about 70 nodes
  • Pile PSD and SBND analysis cluster, with batch, notebooks and Ceph storage
  • Tempest The services cluster carrying identity, registry, OSG and data management
Read more →

Open Data Facility

A United States home for ATLAS Open Data

A mirror of ATLAS Open Data in the US, with notebooks and reproducible workflows attached, so that students, educators and researchers can work with real LHC data without an experiment account.

  • 1 PB Dedicated dCache storage, separate from the production grid pools
  • Runs on RP1 The Lab's multi-site research platform at Chicago and Indiana
  • Notebooks and workflows BinderHub and JupyterHub sessions, plus reproducible REANA workflows
Read more →

The platforms and services we develop on this infrastructure are described under Facilities and Platforms.