Swing

Quick Facts

  • 6x public compute nodes
  • 2x public login nodes
  • 8x NVIDIA A100 GPUS per node
  • 1-2TB DDR4 and 320-640GB GPU memory per node
  • 128 cpu cores per compute node
  • Infiniband HDR Interconnect

Available Queues

Swing has several queues defined. Use the -q option with qsub to select a queue. The default queue is gpu.

Nodes allow for multiple jobs from multiple users up until the resources are fully consumed (8 jobs with 1 GPU each per node, 1 job with 8 GPU per node, and everything in between).

SWING QUEUE Name Number
of
Nodes		 GPUs
Per
Node		 GPU
Memory
Per
Node		 CPUs
Per
Node		 DDR4
Memory
Per
Node		 Local
Scratch
Disk		 Operating
System
gpu 5 8x NVIDIA A100 40GB 320GB 2x AMD EPYC 7742 64-Core Processor (128 Total Cores) 1TB 14TB Ubuntu 22.04 LTS
gpu-large 1 8x NVIDIA A100 80GB 640GB 2x AMD EPYC 7742 64-Core Processor (128 Total Cores) 2TB 28TB Ubuntu 22.04 LTS

File Storage

On Swing, users that want to take advantage of local scratch space will have the option of using a small scratch space on the node’s memory (located at /scratch, 20GB tmpfs). Otherwise, users have access to the same GPFS filesystems as on our other resources including home, project and group space.

Please see our detailed description of the file storage used in LCRC here.

Architecture

Swing runs with 2x AMD EPYC 7742 64-Core Processor and 8x NVIDIA A100 GPUS per node.

Swing is also using an Infiniband HDR interconnect for its network. This fact comes into play when considering MPI programs that would use Infiniband library as a means for communication.

Running Jobs on Swing

For detailed information on how to run jobs on Swing, you can follow our documentation by clicking here: Running Jobs on Swing.

With an eye towards future alignment with the ALCF, LCRC has adopted PBS Pro for the Swing cluster.