Google Compute Engine (GCE)
Google Cloud's virtual machines — live migration, sustained-use discounts, and per-second billing distinguish it from the EC2 model.
What is GCE?
Google Compute Engine (GCE) is Google Cloud’s virtual machine service — the EC2 equivalent, and the compute foundation of GCP. It offers the expected machine families (general, compute-, memory-, accelerator-optimized) but distinguishes itself on a few design choices that reflect Google’s operational philosophy.
The differentiators worth knowing
Live migration: GCE can move a running VM to another host during maintenance without rebooting it — a genuinely different reliability posture from providers that require instance reboots for host maintenance, and a meaningful convenience for long-running stateful VMs. Sustained-use discounts: run an instance a large fraction of the month and GCE automatically discounts it — no reservation commitment required, unlike the Reserved Instance/Savings Plan up-front commitment model (Committed Use Discounts exist too for deeper savings). Per-second billing and custom machine types (pick arbitrary vCPU/memory ratios rather than fixed shapes) let you fit the instance to the workload instead of rounding up to a predefined size — a real cost lever for oddly-shaped workloads. Preemptible/Spot VMs mirror EC2 Spot for interruptible discount compute.
Choosing and sizing
GCE sits under GKE nodes, managed instance groups (Google’s autoscaling construct), and standalone VMs. The optimization playbook rhymes with EC2’s — right-size aggressively (custom types help), lean on sustained-use/committed discounts for steady load, use Spot for interruptible work — with the twist that automatic sustained-use discounts reward simply running steadily without the reservation paperwork, lowering the activation energy for savings. As everywhere, if the workload is a stateless container, Cloud Run may be the better-fitting abstraction than a raw VM.
What people get wrong
- Rounding up to fixed shapes when custom machine types would fit the workload and cut cost.
- Ignoring automatic sustained-use discounts in cost models — GCP discounts steady VMs without a commitment.
- Reflexively applying EC2 mental models — live migration, per-second billing, and custom types change the optimization calculus.
Primary source: Google Compute Engine documentation
Historical figures and technical concepts for informational purposes only. Not technical, professional, legal, or financial advice. Sources: Official Documentation.