1. Introduction to Workload Partitioning

In modern cloud-native architectures, especially at the edge, resource efficiency and system stability are paramount. Workload Partitioning is a sophisticated feature in OpenShift designed to provide strict CPU core isolation between critical OpenShift system components and user-deployed workloads.

This isolation is particularly critical in resource-constrained environments such as:

  • Single-Node OpenShift (SNO): Where the entire control plane and data plane reside on a single machine.

  • 3-node Compact Clusters: Where master and worker roles are co-located.

  • Edge Computing: Where hardware resources are limited and predictable latency is required.

In these scenarios, system services and user applications share the same physical host and compete for CPU resources. Without partitioning, a sudden spike in a user application’s CPU usage could starve critical system processes like etcd or the kubelet, leading to cluster instability or even node failure.

By dedicating specific CPU cores to the system (Reserved) and others to workloads (Isolated), you can:

  • Achieve Predictable Performance: Ensure user applications have guaranteed access to their assigned cores.

  • Ensure Cluster Stability: Protect the control plane from being impacted by user workloads.

  • Minimize "Noisy Neighbor" Effects: Prevent inter-process interference at the hardware level.

This lab provides a comprehensive, step-by-step demonstration of how to enable, configure, and verify workload partitioning. Throughout the session, we will:

  1. Apply a PerformanceProfile to define CPU sets.

  2. Deploy and analyze various Pod types (BestEffort, Burstable, Guaranteed) to see how they are pinned.

  3. Deploy a Virtual Machine using OpenShift Virtualization to verify that virtualization workloads also respect the partitioning rules.

  4. Perform a deep dive into the underlying kernel and runtime configurations.

For official documentation and deep-dive technical details, please refer to: