On Google Cloud! ☁️
The goal of these experiments is to test the Flux Operator at small scales, meaning taking simple workflows and comparing execution between using the Flux Operator and a more native approach.
- lammps
- osu-benchmarks
- resources: getting more than one flux pod per node working!
- indexed-job-timing: comparison of pods a la-carte vs with indexed jobs.
- autoscale: early autoscaling work
- bare-metal-comparison: attempted start to compare GKE to CE, never went far because we couldn't get compact on CE, and lammps to run on GKE.
- networking and service-timing: testing a networking bug with a service
- rootless-kubernetes
- scheduler: testing fluence against the default scheduler
- workflows
For all of these examples, you will need a Google Cloud Project and to enable the Kubernetes Engine API. If you are running these are your personal cloud account, it's recommended to use the pricing calculator before creating any resources!