• Home
      • |
    • Engineering Compute Cluster – Overview

Engineering Compute Cluster – Overview

The Bagley College of Engineering cluster is a shared high-performance computing (HPC) system for research and coursework that needs more processing power than a desktop or laptop can provide. Work is submitted to the cluster through the Slurm job scheduler, and the cluster runs it on available compute nodes.

At a Glance

  • Login hostname: engr-cluster-01.engr.msstate.edu
  • CPU nodes: 20 (24 cores and about 120 GB of memory each, 480 cores total)
  • GPU node: 1 (one NVIDIA A100 80 GB GPU)
  • Operating system: Rocky Linux 9
  • Scheduler: Slurm
  • Access: SSH, from any MSU network or through the MSU VPN

Hardware

CPU nodes

Each CPU node has two Intel Xeon Silver 4310 processors (2.1 GHz) with 12 cores each, for 24 cores per node, and about 120 GB of memory. The nodes are named node1 through node20.

GPU node

The GPU node (gpu1) has the same CPU and memory configuration as the CPU nodes and adds one NVIDIA A100 GPU with 80 GB of memory.

Partitions (Queues)

A partition is a group of nodes that you choose when you submit a job. Every job must name a partition.

Partition Nodes Priority Who can use it
sbatch (default) All 20 CPU nodes (node1 to node20) Low All cluster users
isecompute node1 to node10 High Industrial and Systems Engineering (ISE) only
gpu gpu1 (A100 80 GB) Standard All cluster users

Please note: The sbatch and isecompute partitions share nodes 1 to 10. Because isecompute has higher priority, a job running in the sbatch partition on those nodes can be paused while an isecompute job uses the same resources, and it resumes afterward.

Software

Applications are installed under /usr/local and are managed with environment modules. Load a module to make that software available in your session.

module avail          # list available software
module load ansys/25.2   # load a module
module list           # show what is currently loaded

Currently available modules include:

  • ANSYS (versions 24.2 and 25.2)
  • Abaqus 2023
  • OpenFOAM 2412
  • NVIDIA HPC SDK 23.9 (compilers and MPI variants)
  • CUDA 12.3

The system also provides GCC 11, Python 3.9, and Podman for containers. If you need software that is not listed, contact us (see below).

Storage

All nodes share the same storage, so your files are visible on the login node and every compute node. The compute nodes have no local scratch disk.

  • Home directory: stored on the shared data volume. It is removed when your account is closed.
  • Scratch (/data/scratch): temporary working space that is cleaned between semesters.
  • No backups: home directories and the data volume are not backed up, so keep copies of important files elsewhere.
  • No enforced quotas: please be considerate of other users and clean up files you no longer need.

See Engineering Compute Cluster – Storage for details and file transfer instructions.

Getting Started

  1. Request access to the cluster
  2. Run your first job

Help

For questions or problems with the cluster, email support@engr.msstate.edu. Please include your NetID, a description of the problem, and, if it involves a job, the job ID and any error messages.