Kubernetes

What Is Kubernetes?

Kubernetes – often abbreviated as K8s – is an open-source container orchestration platform originally developed by Google and released as an open-source project in 2014. Today, Kubernetes is governed by the Cloud Native Computing Foundation (CNCF) and is the undisputed standard for running containerised applications in production.

For mid-market companies, Kubernetes solves a concrete problem: how do you reliably run dozens or hundreds of containers at scale without manual intervention? Kubernetes automates deployment, scaling, load balancing, and self-healing of container workloads.

Understanding Kubernetes Architecture

Kubernetes architecture is based on a cluster model with two main components:

• Control Plane (Master): Manages the cluster. Contains the API server, scheduler, controller manager, and etcd (the distributed database for cluster state).
• Worker Nodes: Run the actual workloads. Each node runs a kubelet (agent) and a container runtime such as containerd.

Pods, Nodes, and Clusters

The smallest deployable unit in Kubernetes is a Pod – one or more containers sharing network and storage. Pods run on Nodes (physical or virtual machines). Multiple Nodes form a Cluster. In practice, you define the desired state of your application declaratively in YAML manifests, and Kubernetes ensures that state is maintained.

For example: you define that your web shop should run with three replicas. If a Pod fails, Kubernetes automatically starts a new one. If load increases, a Horizontal Pod Autoscaler can add more Pods.

Kubernetes vs. Docker Swarm

Docker Swarm is Docker's native orchestration solution and is significantly simpler to set up than Kubernetes. For small teams with few containers, Swarm may suffice. However, Kubernetes offers decisive advantages:

• Scalability: Kubernetes scales to thousands of nodes and tens of thousands of pods. Docker Swarm reaches its limits sooner.
• Ecosystem: Kubernetes has a vast ecosystem with Helm Charts, Operators, Service Meshes, and more.
• Community and Support: All major cloud providers offer Managed Kubernetes. Docker Swarm has significantly less community support.
• Self-Healing: Kubernetes detects failed containers and replaces them automatically – more reliably than Swarm.

When Is Docker Swarm Sufficient?

For a small development team with fewer than ten containers and no complex scaling requirements, Docker Swarm can be a pragmatic start. However, once you need multiple environments (staging, production), auto-scaling, or a complex deployment model, Kubernetes is the better choice.

Managed Kubernetes vs. Self-Hosted

For mid-market companies, the question "Managed or Self-Hosted?" is often decisive:

• Managed Kubernetes (EKS, AKS, GKE): The cloud provider operates the control plane, handles updates, and ensures high availability. You pay a management fee (e.g., approximately $73/month per cluster for AWS EKS) but save significantly on operational effort.
• Self-Hosted: You operate the entire infrastructure yourself – including the control plane. Full control but high effort for updates, security, and availability. Requires at least one to two dedicated Kubernetes administrators.

Recommendation for the Mittelstand

For most mid-market companies, we recommend Managed Kubernetes. The operational overhead for self-hosted clusters is regularly underestimated. With EKS or AKS, you can focus on your applications rather than maintaining cluster infrastructure. The additional cost for the managed service is typically far lower than the staffing costs for dedicated Kubernetes administrators.

Kubernetes Cost Overview

Kubernetes costs consist of:

• Compute: Worker Nodes (EC2 instances, VMs) are the largest cost component.
• Managed Service Fee: Approximately $73/month per cluster for AWS EKS.
• Networking: Load balancers, NAT gateways, data transfer.
• Storage: Persistent volumes for databases and stateful workloads.

A typical production cluster for a mid-market company with three Worker Nodes costs between $500 and $2,000 per month – depending on instance type and workload. With Spot Instances and right-sizing, these costs can be reduced by 30–60%.