nexus, index source, nexus main

high performance computing

We oversimplify High Performance Computing to mean: Multiple computers networked together with very low communication latency creating in effect a supercomputer. How is this done on prem? How is it done on the three primary cloud platforms?

uw on-premise hpc

This will be a slightly abstracted description of working on facilities at the University of Washington.

• Get white listed with access to Klone, Tillicum and Kopah
  • Respectively HPC blades, GPU sysytems, and S3-style object storage
  • Follow instructions to get MFA
• ssh NetID@klone.hyak.uw.edu
  • This is the login node (not necessarily where work gets done)

AWS Parallel Computing Service PCS

• AWS PCS introduction / overview video
• Slurm controller
• Login node group
• Compute node group

Azure

GCP