Sole-Tenant Node

Sole-tenant nodes are dedicated physical servers that host only your VMs, billed for the entire node whether or not VMs fill it. Under-packed sole-tenant nodes waste the premium you pay for dedicated hardware.

Paying for the whole server, occupied or not

The meter on a sole-tenant node is the node: an entire physical host’s worth of vCPUs and memory, charged for every hour the node is reserved to you. The VMs placed on it do not add compute charges; they consume capacity you have already bought. That inverts the usual cloud economics. On multi-tenant Compute Engine, an idle VM stopped is money saved; on a sole-tenant node, stopping a guest VM changes the bill not at all, because the host underneath keeps billing at full price.

Density is the only cost lever

Since the node’s price is fixed, cost efficiency is purely a function of how much of it you use. ZopDev discovers sole-tenant nodes via Cloud Asset Inventory and analyzes VM packing density to flag under-utilized nodes: hosts where the guest VMs’ combined footprint claims a small slice of the hardware you are paying to reserve exclusively. The finding is always the same shape: consolidate guests onto fewer nodes, or shrink the node group.

Where the dedicated-hardware premium leaks

Common cases: a node group sized for a migration wave that ended, leaving half-empty hosts; workloads that justified sole tenancy for licensing reasons being decommissioned while their nodes remain; and autoscaled node groups whose minimum keeps a spare host reserved around the clock for guests that arrive twice a month. Each is invisible from the VM list (the guests all look healthy) and only visible from the node’s occupancy.

Node groups and occupancy in the console

Google Cloud console → Compute Engine → Sole-tenant nodes lists node groups, their node templates, and per-node occupancy, meaning which VMs sit on which host. Nodes showing a short guest list against a large template are the ones the packing-density analysis will have flagged.