|---|---|
| **Pod** | Smallest deployable unit (one or more containers) |
| **Node** | Physical or virtual machine running pods |
| **Cluster** | Multiple nodes managed as a single k8s system |
| **Deployment** | Defines how many pod replicas to run and how to update them |
| **Service** | Network abstraction that exposes pods |
| **Ingress** | Routes external HTTP traffic to services |
| **HPA** | Horizontal Pod Autoscaler — scales pod count based on load | ## Docker vs Kubernetes — Head to Head | Item | Docker + Compose | Kubernetes |
|---|---|---|
| **Role** | Build and run containers | Orchestrate containers at scale |
| **Server scale** | 1–3 servers | 3+, typically 10+ |
| **Learning curve** | Low (1–2 weeks) | High (3–6 months of hands-on experience) |
| **Configuration** | docker-compose.yml (dozens of lines) | YAML files — hundreds to thousands of lines |
| **Auto-scaling** | Manual or limited | Fully automatic (HPA) |
| **Zero-downtime deploy** | Manual implementation | Built-in (Rolling Update) |
| **Cloud cost** | Server cost only | Cluster management fee (GKE: $73+/month minimum) |
| **Ideal team size** | 1–5 people | 5+ with a DevOps engineer | ## Solo Developer Decision Tree ```
Q1. How many servers do you need? → 1–2 servers: Docker Compose is enough → 3+ servers: consider k8s or PaaS Q2. Do you need to handle 10× traffic spikes? → No: Docker Compose → Yes: PaaS (Vercel/Railway) or k8s Q3. Do you need 99.9%+ uptime (< 8 hours downtime/year)? → No: Docker Compose + monitoring → Yes: k8s or managed k8s (GKE/EKS/AKS) Q4. Do you have a dedicated DevOps person? → No (solo dev): try PaaS first → Yes (team): consider direct k8s

gcloud run deploy myapp \ --image gcr.io/myproject/myapp:latest \ --platform managed \ --allow-unauthenticated \ --max-instances 10
# Cost: ~$0.24 for 1M requests + 1vCPU/hour; $0 when idle