vCPU versus pCPU and time slots

The core of the vSphere system is the so called hypervisor—a process that serves running virtual machines. One of the main tasks of a hypervisor is to manage a VM's access to hardware resources in the most efficient way. It does this according to the available resources and configures the importance of running VMs.

All vCPUs assigned to a VM are treated as a chain. Each vCPU is a link in this chain. In vSphere terms, it's called world. Each VM has as many worlds as it has vCPUs assigned plus two invisible service worlds. One of these service worlds is responsible for the VM itself; the other one is responsible for its input/output.

All such chains from each VM create a queue, which becomes an input to the hypervisor process. Each VM gets a timeslot—a period of time when a hypervisor gives it access to the actual hardware. During its timeslot, a VM gets access to a host's physical CPU (pCPU). The timeslot is limited to less than 1 millisecond. Once it's over...