A vCPU is a software-defined unit of processing power assigned to a virtual machine. It does not exist as a physical component but represents a portion of the physical CPU (pCPU) resources managed by the hypervisor's scheduling engine.
Execution Mechanism
The hypervisor divides physical CPU cycles into time slices and allocates them to vCPUs. On systems with Hyper-Threading enabled, a vCPU typically maps to a single logical thread. This abstraction allows multiple virtual instances to share the same hardware concurrently, maximizing the utilization of the physical silicon.
- Granular Scaling: Administrators can provision precise compute capacity for specific workloads.
- Resource Pooling: Physical cores are treated as a shared pool for all virtual instances.
- Dynamic Allocation: vCPU counts can be adjusted via software without hardware intervention.
In high-density environments, the overcommit ratio is a key performance factor. For instance, a host with 32 logical threads might support 96 vCPUs at a 3:1 ratio. Performance degradation occurs when the CPU Ready metric rises, indicating that vCPUs are waiting for physical execution time. Keeping this latency low is essential for real-time applications and database performance.