Compute Autoscaling
| Service | Min instances | Min instances for HA | How? | Speed |
|---|
| Compute Engine | 1 instance | 2 instances | MIG / MIB + load balancer | Moderate / better |
| GKE Pods | 1 Pod | 2 Pods (e.g. Deployment) | Load-based—Horizontal Pod Autoscaler | Decent |
| GKE Node Pools | 1 node | 3 nodes | Kubernetes Pod back-pressure | Slow |
| App Engine Flexible | 1 container | 2 containers | Request-based (same autoscaler as Compute Engine) | Gradual |
| App Engine Standard | 0 | 0 | Request-based | Almost instantaneous |
| Cloud Run | 0 | 0 (adjustable) | Request-based | Very fast |
| Cloud Functions | 0 | 0 (optional) | Request-based | Almost instantaneous |