HPC Workflow Simulator

This demo illustrates how research jobs traverse a Slurm cluster:

• Internet cloud links your laptop to the campus Login node.
• The Login and Head nodes communicate over Ethernet.
• An InfiniBand switch (rectangle) fans out to 10 compute nodes and high‑speed GPFS storage.
• Jobs run for 20 seconds per node; multi‑node jobs generate extra cross‑node IB traffic.
• Up to 20 jobs may run concurrently; extras queue until enough free cores are available.
• Hover nodes for live stats (cores, packets); cyan packets show real‑time network flow.

Quick Scenarios to Try

Scenario	Cores	What to Watch
Single‑core “hello world”	1	One compute node blips; minimal GPFS traffic.
Single‑node full load	64	One node hits 100 % utilisation.
Two‑node MPI test	128	Cross‑node IB packets between two compute nodes.
One rack (5 nodes)	320	Half the cluster active, IB switch busy.
Cluster‑wide flood	640	All compute nodes + GPFS at 100 %.
Queue pressure	25 jobs × 32	20 run, 5 wait—observe queue dynamics.

Cores (1 ‑ 640): Submit Job