Sizing virtualization hosts
You have a VM count, a growth guess, and a budget window, and you need to turn that into a number of physical hosts you can defend. The math is not complicated, but the assumptions matter: vCPU overcommit ratio, memory per VM (which almost always binds before CPU does), and whether the cluster survives losing a host during maintenance.
The honest starting points: a 4:1 vCPU-to-core ratio is a safe default for mixed workloads, memory should not be overcommitted for production, and N+1 is not optional if the estate matters. Databases and latency-sensitive workloads push toward 2:1 or 3:1; light dev/test estates tolerate 6:1 or more.
What actually determines the answer
| Factor | Why it matters |
|---|---|
| Memory per VM | Memory binds before CPU in most estates. Size RAM first, then check cores. |
| vCPU overcommit ratio | 4:1 for mixed workloads, tighter for databases, looser for dev/test. This single assumption moves host count more than any other. |
| N+1 headroom | One host must be able to fail (or be patched) with the cluster still holding its load. |
| Vendor standardization | Matching your existing tooling (iDRAC vs iLO, existing spares) usually beats a marginal spec difference. |
The paths, with their trade-offs
Dell PowerEdge R770 hosts
When your estate is Dell-standardized or you're starting fresh on the current Intel platform.
Trade-off: A newer platform means a longer runway but less field history than 16G.
Configured BOM
R770 - 60 VM virtualization host
- CPU 2× Intel Xeon 6 processor, 48 cores each
- Memory 8× 64 GB DDR5 RDIMM (512 GB total)
- Storage 6× 3.84 TB NVMe read-intensive SSD
- Network Dual-port 100 GbE adapter
- Support 4-year ProSupport Plus, next business day
HPE ProLiant DL380 Gen12 hosts
When your estate runs on HPE tooling and support relationships.
Trade-off: Functionally equivalent to the R770 path; the deciding factor is your operations stack, not the spec sheet.
Configured BOM
DL380 Gen12 - 60 VM virtualization host
- CPU 2× Intel Xeon 6 processor, 48 cores each
- Memory 8× 64 GB DDR5 RDIMM (512 GB total)
- Storage 6× 3.84 TB NVMe read-intensive SSD
- Network Dual-port 100 GbE adapter
- Support 4-year HPE support, next business day
Smaller hosts, more of them
When failure-domain size worries you more than rack space.
Trade-off: More hosts means more licensing and more switch ports; the per-VM cost usually rises.
Configured BOM
R770 - 30 VM virtualization host
- CPU 2× Intel Xeon 6 processor, 32 cores each
- Memory 8× 32 GB DDR5 RDIMM (256 GB total)
- Storage 4× 1.92 TB NVMe read-intensive SSD
- Network Dual-port 25 GbE adapter
- Support 4-year ProSupport Plus, next business day
Run your own numbers first: VM sizing calculator →
What deployment involves
Typical scope: rack and cable, firmware baseline, hypervisor install, cluster configuration, and workload migration in maintenance windows.
A refresh of a small cluster is usually measured in weeks, not months; the long pole is migration windows, not hardware.