Kubernetes Architecture Explained
Plain-language K8s with TCO reality check.
Kubernetes Architecture for Decision Makers
Kubernetes has become the de facto standard for container orchestration, but understanding its architecture is essential before committing to the platform. This guide breaks down K8s components in plain language and provides a realistic TCO assessment for SMBs.
Control Plane Components
The control plane is the brain of your Kubernetes cluster:
- kube-apiserver — The front door. Every kubectl command, every pod scheduling decision, and every service discovery lookup goes through the API server. It's the only component that talks to etcd.
- etcd — The cluster's database. Stores all configuration and state as key-value pairs. Losing etcd means losing your cluster.
- kube-scheduler — Decides which node runs each pod based on resource requests, affinity rules, taints, and tolerations.
- kube-controller-manager — Runs reconciliation loops that ensure actual state matches desired state (e.g., maintaining replica counts).
Worker Node Components
- kubelet — The agent on each node that receives pod specs from the API server and ensures containers are running.
- kube-proxy — Manages network rules for Service routing. Implements iptables or IPVS rules for load balancing.
- Container runtime — Actually runs containers. containerd is the standard runtime (Docker was deprecated as a runtime in K8s 1.24).
How a Pod Gets Scheduled
- You apply a Deployment manifest via
kubectl apply - The API server validates and stores it in etcd
- The Deployment controller creates a ReplicaSet
- The ReplicaSet controller creates Pod objects
- The scheduler assigns each Pod to a node
- The kubelet on that node pulls the image and starts the container
- kube-proxy updates iptables rules for Service discovery
Managed vs Self-Managed Kubernetes
| Factor | Managed (EKS/GKE/AKS) | Self-Managed (kubeadm) |
|---|---|---|
| Control plane cost | $72-150/month | 3 VMs (~$150-300/month) |
| Upgrades | One-click | Manual, risky |
| etcd management | Handled by provider | Your responsibility |
| SLA | 99.95% | Depends on your team |
| Customization | Limited | Full control |
TCO Reality Check for SMBs
A realistic Kubernetes cost breakdown for a small production workload:
- EKS control plane: $73/month
- 3 worker nodes (t3.medium): ~$100/month
- Load Balancer: ~$20/month
- Monitoring stack: ~$50/month (Prometheus/Grafana)
- Engineering time: 20-40 hours/month for maintenance
Total: ~$250/month + significant engineering investment. For teams running fewer than 10 microservices, consider whether ECS, App Runner, or even a single well-configured EC2 instance might be more cost-effective.
When Kubernetes Makes Sense
- You have 10+ microservices
- You need multi-cloud portability
- Your team has Kubernetes expertise (or budget to build it)
- You need advanced deployment strategies (canary, blue-green)
- Your workloads have variable resource requirements
Eazy SaaS Tip: We help SMBs evaluate whether Kubernetes is the right choice for their workload. In many cases, a simpler architecture saves 60% on infrastructure costs while being easier to operate.