Cisco
Login
Search
Software Secure Workload
Activity Configure

Manage Policy Lifecycle in Secure Workload Grouping Workloads: Clusters and Inventory Filters Clusters Cluster Confidence

Last updated: Jun 09, 2025

Cluster Confidence

Use the confidence or quality score of a cluster to identify clusters needing improvement.

The confidence for a cluster is the average of the confidences for member workloads. In general, the more similar a workload is to other members of the cluster it was assigned, and the more dissimilar it is to the workloads of the closest (most similar) alternative cluster, the higher the confidence for that workload.

When flows are used for clustering, two workloads are similar when they have a similar pattern of conversations (such as similar sets of neighbors in the conversation graph, i.e., similar sets of consumer and provider workloads and ports).


 

  • Cluster confidence is not computed (undefined) for:

    • clusters containing only one workload

    • approved clusters

    • workloads in the scope for which no communication was observed (or no process information is available, if process-based clustering was chosen)

  • Clusters do not span partition boundaries (such as subnet boundaries, see route labels in the advanced automatic policy discovery configurations). However, in computing confidence and alternate cluster, such boundaries are ignored. This indicates the potential existence of workloads or clusters that behave very similarly even though they are in different subnets.

  • After editing clusters, the confidence scores may become inaccurate as they are NOT recomputed until you discover policies again.

To view cluster confidence, see View Clusters.

Previous topic Clusters Next topic View Clusters