Check your vCPU to physical CPU ratio against workload-appropriate limits and read it against observed CPU ready time.
The ratio itself is simple division, but it is only half the picture: ESXi will happily schedule 20:1 and the scheduler decides whether that hurts. CPU ready time is the empirical answer — under 3% the ratio is fine whatever it says, and above 5% it is too high even if the arithmetic looks conservative. A ratio chosen from a spreadsheet without checking ready time either wastes hardware or produces mystery latency that application owners blame on the storage array.
VMware vCPU Overcommit
ratio = total allocated vCPUs ÷ (hosts × cores × threads per core); the ratio is then compared with the profile ceiling and cross-checked against CPU ready time.
ratio = total allocated vCPUs ÷ (hosts × cores × threads per core); the ratio is then compared with the profile ceiling and cross-checked against CPU ready time. The ratio itself is simple division, but it is only half the picture: ESXi will happily schedule 20:1 and the scheduler decides whether that hurts. CPU ready time is the empirical answer — under 3% the ratio is fine whatever it says, and above 5% it is too high even if the arithmetic looks conservative.
A ratio chosen from a spreadsheet without checking ready time either wastes hardware or produces mystery latency that application owners blame on the storage array.
This calculator takes 6 inputs: Hosts in cluster, Physical cores per host, Count hyper-threads as logical CPUs, Total vCPUs allocated, Workload profile, Observed CPU ready time. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
Yes — for latency-sensitive VMs such as trading engines, real-time telco workloads or licensed database cores, where you also pin CPU affinity and set full reservations. Everywhere else 1:1 wastes most of the hardware you bought.