Compute Architecture
OSC provides compute capacity for Kubernetes worker nodes from dedicated customer machine pools. This means that worker workloads from different customers are not placed on the same physical compute pool.
The physical servers are operated by OSC. They run the virtualization and networking components required to create worker machines for Shoot clusters. Customers do not manage hypervisors, operating-system images or bare-metal provisioning. These functions are part of the platform.
Within each Availability Zone, compute nodes are organized into local virtualization clusters. A customer machine pool spans the region through separate zone-local clusters. This keeps failures and maintenance activities contained while still allowing a Shoot to use workers in multiple zones.
Gardener requests worker machines through the OSC infrastructure layer. The platform selects capacity from the appropriate customer pool, creates the machine, attaches networking and storage, and joins it to the Shoot cluster.
The dedicated model provides a strong operational and isolation boundary, but capacity is still finite. The maximum number and size of worker nodes depend on the installed physical resources.
Customers should plan expected growth, peak demand and maintenance headroom together with OSC.
From the customer perspective, the compute abstraction remains a Kubernetes worker pool. The complexity of physical hosts, virtualization clusters and machine lifecycle is intentionally hidden behind the managed service.