Provisioning clusters

Wisp uses the term "cluster" for instances and nodes interchangably. For Wisp, a cluster is simply a resource the can execute code using whatever hardware it has available.

A cluster can be an instance from any of the supported cloud providers, but in the future, also a cluster of instances or even a serverless endpoint. Wisp only cares about cluster being able to solve the workload as efficiently as possible.

Clusters are inherently attached to Jobs, as the job decides what requirements the resources should optimize for. For that reason, you can only provision clusters by using wisp run based on your wisp-spec.yml file.

You can however manually start stop or destroy existing clusters with Wisp, both from the CLI or from the Dashboard. Go to the next page to see how.

Billing

Wisp only bills you for running resources. That means that a stopped cluster will not cost you anything. As soon as you restart it, Wisp will resume billing. You can always see how much you're being charged in the dashboard under Admin -> Cost Center.