VM Density Calculator

How many virtual machines a host can hold, and whether CPU or RAM is the limit.

This is a planning estimate, not a measurement or a guarantee. It does not model NUMA topology, workload bursts, storage or network bottlenecks, HA and failover reserve, hypervisor scheduling behaviour, memory ballooning or page sharing, licensing constraints, or heterogeneous VM sizes. Validate any density target against measured utilisation before committing to it.

What does this calculator do?

VM density is set by whichever resource runs out first. This calculator works out both limits separately — how many VMs the CPU allows once the oversubscription ratio is applied, and how many the memory allows after the hypervisor’s reservation — and reports the smaller figure as the capacity, along with which resource is the bottleneck and how much RAM is left stranded. Seeing the bottleneck is the point: adding cores to a memory-bound host buys nothing. Memory is not oversubscribed in this model, which matches the conservative practice of sizing RAM one-to-one. Everything is computed in your browser from the values you type. Nothing is looked up on the network — there is no ping, no DNS query, no WHOIS and no geolocation.

Formula

Two independent limits, and the capacity is the lower of them.

cpuCapacity = physicalCores × oversubscriptionRatio
cpuVmCount  = floor(cpuCapacity ÷ vmVcpu)

usableRam   = hostRam − reservedRam
ramVmCount  = floor(usableRam ÷ vmRam)

vmCapacity  = min(cpuVmCount, ramVmCount)

How to use it

Enter the host’s physical core count — cores, not threads, unless your sizing standard explicitly counts hyper-threads. Enter total host RAM and the amount reserved for the hypervisor and its services. Then describe the typical VM: its vCPU count and its RAM. Finally set the CPU oversubscription ratio your operations standard allows: 1:1 for latency-sensitive workloads, 3:1 to 5:1 for general server VMs, higher for idle test and development estates. Remember to leave capacity for cluster failover on top of this figure.

Worked example

A host with 32 cores and 512 GB of RAM, 32 GB reserved, running VMs of 4 vCPU and 16 GB at a 4:1 ratio. CPU: 32 × 4 = 128 vCPU ÷ 4 = 32 VMs. RAM: (512 − 32) = 480 GB ÷ 16 = 30 VMs. Capacity is 30 VMs and RAM is the bottleneck — 8 vCPU of CPU capacity is left with no memory to go with it.

Formula version1.0.0
FormulacpuCapacity = physicalCores × oversubscriptionRatio cpuVmCount = floor(cpuCapacity ÷ vmVcpu) usableRam = hostRam − reservedRam ramVmCount = floor(usableRam ÷ vmRam) vmCapacity = min(cpuVmCount, ramVmCount)

Frequently asked questions

Why is memory not oversubscribed here?

Because relying on it is risky. Ballooning, page sharing and swap can overcommit memory, but when the guests actually use it the host swaps and performance collapses. Sizing RAM one-to-one is the conservative default this model uses.

What oversubscription ratio is safe?

It depends entirely on the workload. Latency-sensitive databases and VDI often stay at 1:1 or 2:1; general server VMs commonly run 3:1 to 5:1; idle lab estates go far higher. There is no universal number, so enter your own standard.

Does this include N+1 failover capacity?

No. It gives the density of a single host at full utilisation. A cluster that must survive a host failure has to keep one host’s worth of capacity free, so plan the cluster total accordingly.

Sources

Related calculators

Last updated: 2026-08-10 · v1.0.0 · Report an error