Execution Policy & CPU Share Factor
Control where STRATUM places workloads, whether the controller participates in execution, and how CPU capacity is shared across eligible Continuum workers.
Keep the controller out of workload placement
STRATUM can use the controller as an execution host, but larger or production deployments often benefit from keeping the controller focused on orchestration and Fabric services. Select Disable workload execution on controller to exclude the controller from new workload placement while leaving the controller itself, Network Fabric, and already-running workloads intact.
This is not the same as disabling Network Fabric and it does not tear down the controller. It simply removes the controller from the scheduler's eligible placement pool for new workload execution. In the example below, the controller is excluded and two Continuum workers remain available for placement.
CPU Share Factor
CPU Share Factor controls how aggressively STRATUM may share physical CPU capacity across eligible execution workers. The default is appropriate for most environments. Increase sharing only when the workload mix and operational policy intentionally permit CPU oversubscription.
Changes affect new placement decisions. Existing workloads are not automatically stopped merely because the factor is changed.
Execution CPU Share Factor
Use the Execution administration view to select the CPU Share Factor and review the workers that remain eligible for placement.
Disable controller workload execution
- Open Administration → Execution.
- Review the Worker Nodes count and worker table so you know which Continuum workers are available.
- Select Disable workload execution on controller.
- Save/apply the setting if prompted by the current UI.
- Confirm the green status message reports controller workload execution disabled.
- Deploy a new workload and verify placement selects an eligible worker rather than the controller.
Separate control-plane responsibility from workload pressure
This policy lets a small STRATUM installation use every node when desired, while allowing a scaled deployment to reserve the controller for coordination, APIs, Fabric state, authentication, scheduling decisions, and other platform duties. The execution workers can then absorb workload CPU, memory and GPU pressure without turning the controller into an accidental capacity target.
The outcome you should see
The controller remains online and fully participates in STRATUM control/Fabric operations, but new workloads are placed only on eligible execution workers. CPU Share Factor continues to define how those execution workers expose CPU capacity to placement.
