Compute and clusters
A cluster is a Kubernetes or Slurm cluster Vantage manages on your behalf. Compute profiles describe the shape of compute (GPU type, count, autoscaling bounds, instance class) and are reused across products. Compute providers (AWS, Azure, GCP, on-prem) are the physical substrate.
Cluster types
Vantage provisions three kinds of clusters, Slurm, Kubernetes, and Slurm on Kubernetes. Your choice determines the scheduler and which workloads can run. For the comparison table and the status lifecycle, see Clusters concepts.
Compute profiles
A compute profile bundles GPU vendor and count per node, instance type, autoscaling min/max, and a default cost rate. Sessions, training jobs, sweeps, and endpoints all reference a profile, choose one once and the rest of the form configures around it.
Profiles live at the workspace level. Your admin defines which profiles are available.
Compute providers
Providers are the physical infrastructure Vantage provisions clusters on, public clouds, on-premises, and partner colocation. For the provider comparison table and validity lifecycle, see Clusters concepts and Compute providers concepts.
Regions and availability
Cloud clusters run in the region you select during creation. Slurm clusters can span multiple availability zones within a region. On-premises clusters report their location as configured by your admin.
Cross-references
- Clusters concepts: cluster types, status lifecycle, partitions and compute pools, cost
- Compute providers concepts: cloud accounts, provider types, validity