Work out how many hypervisor hosts replace a fleet of physical servers, and the capex and power that consolidation changes.
Consolidation works because standalone servers sit almost idle: sizing against measured consumption rather than installed capacity is what produces ratios of 15:1 or better. CPU can then be overcommitted, but memory cannot, so the memory side usually decides the host count while the CPU side decides whether the overcommit ratio is defensible. Consolidation business cases live or die on the ratio, and sizing from installed cores instead of consumed cores understates the ratio by an order of magnitude — while ignoring the memory constraint overstates it. Treat the capex and electricity figures as editable placeholders and substitute your own quotes and tariff.
Hypervisor Consolidation
hosts = max(consumed cores ÷ (host cores × overcommit × target utilisation), consumed memory ÷ (host RAM × target utilisation × 0.94)), rounded up.
hosts = max(consumed cores ÷ (host cores × overcommit × target utilisation), consumed memory ÷ (host RAM × target utilisation × 0.94)), rounded up. Consolidation works because standalone servers sit almost idle: sizing against measured consumption rather than installed capacity is what produces ratios of 15:1 or better. CPU can then be overcommitted, but memory cannot, so the memory side usually decides the host count while the CPU side decides whether the overcommit ratio is defensible.
Consolidation business cases live or die on the ratio, and sizing from installed cores instead of consumed cores understates the ratio by an order of magnitude — while ignoring the memory constraint overstates it. Treat the capex and electricity figures as editable placeholders and substitute your own quotes and tariff.
This calculator takes 13 inputs: Physical servers to consolidate, Cores per physical server, Average CPU utilisation today, Memory per physical server, Average memory utilisation today, Cores per new hypervisor host, Memory per new hypervisor host, Target utilisation on new hosts, vCPU per physical core, Purchase price per hypervisor host, Power draw per physical server, Power draw per hypervisor host, Electricity price. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
A physical server bought for peak load runs at roughly 10% average CPU, so its installed cores are mostly idle capacity you are paying to cool. Virtualisation's whole value is reclaiming that gap, which only appears if you size on measured consumption plus headroom.