Concepts

Three mental models behind compute federations in Vantage.

Federation as a cluster group

A federation groups multiple clusters into a single logical compute pool. Workloads submitted to the federation can run on any cluster within it. The federation doesn't replace individual cluster management — each cluster keeps its own partitions, compute pools, and configuration.

Workload routing

When a job targets a federation instead of a specific cluster, Vantage routes it to an available cluster within the federation. Routing considers cluster capacity, partition availability, and job requirements. If no cluster can accept the job, it waits in the federation queue.

Use cases

Scenario	Why use a federation
Multi-region deployments	Run jobs in the closest region automatically
Burst to cloud	Keep a small on-prem cluster and burst to cloud clusters during peak demand
Provider diversity	Spread workloads across AWS, Azure, and GCP clusters for resilience
Gradual migration	Move workloads from an old cluster to a new one incrementally

Federations are managed entirely through the Vantage CLI. See the CLI reference for all commands.