Score CPU, memory, storage and network utilisation together to find stranded capacity and the annual cost of the waste.
Utilisation is only meaningful alongside balance. An estate at 42% CPU and 78% memory is not half empty — it is memory bound with stranded CPU that cannot be sold, because the next VM needs both. The spread between the highest and lowest resource is therefore a better refresh signal than any single average. Stranded capacity is fully paid for and unusable, so the balance score points directly at what the next hardware refresh should change — more memory per socket rather than more sockets, for instance. Replace the cost figure with your own fully loaded annual spend.
Infrastructure Utilization
overall = CPU × 0.35 + memory × 0.35 + storage × 0.20 + network × 0.10; stranded = target − overall; waste = annual cost × stranded ÷ 100.
overall = CPU × 0.35 + memory × 0.35 + storage × 0.20 + network × 0.10; stranded = target − overall; waste = annual cost × stranded ÷ 100. Utilisation is only meaningful alongside balance. An estate at 42% CPU and 78% memory is not half empty — it is memory bound with stranded CPU that cannot be sold, because the next VM needs both. The spread between the highest and lowest resource is therefore a better refresh signal than any single average.
Stranded capacity is fully paid for and unusable, so the balance score points directly at what the next hardware refresh should change — more memory per socket rather than more sockets, for instance. Replace the cost figure with your own fully loaded annual spend.
This calculator takes 6 inputs: CPU utilisation, Memory utilisation, Storage utilisation, Network utilisation, Target utilisation, Annual infrastructure cost. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
Failover capacity, scheduling headroom and burst absorption all live in the unused fraction. A target of 70-80% is the practical optimum: below it you are wasting money, above it a single host failure or workload spike has nowhere to go.