Depending on the configuration, different types of stressors might end up competing for the same resources. For example, if PAO (Performance Add-on Operator) is used on a cluster, kubelet and openvswitch are pinned to the same CPUs.
Being able to generate an heterogeneous, realistic workload (e.g: combination of network load and k8s control plane tasks) would be a very good way of testing how this resource sharing affect the overall cluster scalability.
Depending on the configuration, different types of stressors might end up competing for the same resources. For example, if PAO (Performance Add-on Operator) is used on a cluster, kubelet and openvswitch are pinned to the same CPUs.
Being able to generate an heterogeneous, realistic workload (e.g: combination of network load and k8s control plane tasks) would be a very good way of testing how this resource sharing affect the overall cluster scalability.