Skip to main content
We do every deployment together with your team. An Expanse engineer works alongside your platform team from the first call until Expanse is running across your cluster.

​
How it goes

1

A first call

We learn how you schedule today, what’s on your cluster and what you’d like to fix. You’ll know by the end whether Expanse fits.
2

Security and deployment review

We share the deployment documentation: every outbound call, what the collectors need on your nodes, and the services Expanse runs on. Your security team reviews it and we answer what it raises.
3

Install

Your platform team installs Expanse with one of our engineers alongside. We adapt the layout to your rules, whether that’s Kubernetes or virtual machines, a connected network or none.
4

A month of predictions

For the first month Expanse predicts what every job will need and records each prediction, but shows no recommendation to anyone. At the end of the month, we check the predictions with you against what your jobs actually used.
5

Roll out

Recommendations go to the teams you pick first, then to the rest of the cluster. Results are measured against your own scheduler’s accounting, so you can check every number yourself.

​
What we’ll need from you

  • A cluster to connect: SLURM, Kubernetes, or the scheduler you’ve built.
  • A Kubernetes namespace inside your network for Expanse’s services, or virtual machines where Kubernetes isn’t an option, with a PostgreSQL database, ClickHouse and object storage. We can install the databases for you.
  • An OIDC client from your identity provider, and the email domains that sign in.
  • An engineer from your platform team to install with us, and a reviewer from your security team.
The deployment documentation lists every item in detail, with sizing, once we’ve spoken.

​
Talk to us

Book a call

expanse.sh/contact