Compute job platform
Slurm jobs from a browser. Researchers sign in with their AU account, pick data from ERDA and submit; a React/Node app sits in front of a cluster that grows and shrinks itself on Oracle Cloud.
- why
- Heavy jobs without learning Slurm, SSH and batch scripts first. Use is bursty, so a fixed cluster is either idle or full.
- interesting
- A queued job has to say why it's waiting while a node boots, instead of just looking stuck.