One way that we can guarantee app performance is to simply schedule more contemporary threads.
However, those threads should have some guaranteed operating headroom. It's annoying if pods are killed by OOM because we didn't specify their requirements in the metadata.
If we set the 'requests' flag in the OCP deployment configuration we'll be reserving resources for the scheduled pods, guaranteeing that if a pod is scheduled, it'll at LEAST have the resources requested.
One way that we can guarantee app performance is to simply schedule more contemporary threads.
However, those threads should have some guaranteed operating headroom. It's annoying if pods are killed by OOM because we didn't specify their requirements in the metadata.
If we set the 'requests' flag in the OCP deployment configuration we'll be reserving resources for the scheduled pods, guaranteeing that if a pod is scheduled, it'll at LEAST have the resources requested.