From a host or cluster perspective, CPU allocation is the ratio of virtual machine CPUs (vCPU) to physical ESXi host cores (pCPU). Ignoring buffers and overcommitment (which will be discussed shortly) allows an administrator to manage capacity simply based on this one to X mapping. Using CPU allocation as a capacity container is usually based on environment design or special use cases. For example, if an administrator had a cluster dedicated to mission-critical applications and there was an associated organizational policy that CPU overcommitment should not exceed 2:1 (2 vCPUs to every pCPU), that administrator could then use the allocation model in combination with overcommitment to ensure that this ratio was never breached.

Recommendation for CPU allocation:

As a general rule for generic production server clusters, it is recommended to either disable this container altogether or ensure that you have an overcommitment ratio high enough so that this container does not become the most constraining resource unnecessarily.