Kubernetes has already been established as the OS of the cloud. As such, it sits between the applications and infrastructure. Kubernetes runs on infrastructure, and applications run on Kubernetes.

Kubernetes on different infrastructure platforms

The illustration shows four Kubernetes installations running on four different infrastructure platforms. Because Kubernetes abstracts the underlying infrastructure, the application at the top of the diagram can run on any of the Kubernetes installations. You can also migrate it from one Kubernetes installation to another.

A Kubernetes installation is known as a Kubernetes cluster.

There are a couple of things worth clarifying here:

It is unusual for a single Kubernetes cluster to span multiple infrastructures.

For example, you likely won’t see Kubernetes clusters that span multiple clouds. Likewise, you’ll rarely see clusters that span on-prem and the public cloud. This is mainly due to network speed and reliability. Generally speaking, you want high-speed reliable networks connecting the Nodes in a cluster.

Although Kubernetes can run on many platforms, applications that run on Kubernetes have stricter requirements.

You will learn about this later on in the chapter, but Windows applications will only run on Kubernetes clusters with Windows Nodes, Linux applications only run on clusters with Linux Nodes, and applications written for ARM/Raspberry Pis require clusters with ARM Nodes. 

A Kubernetes cluster consists of one or more machines that have Kubernetes installed on them. The machines can be physical servers, virtual machines (VM), cloud instances, your laptop, Raspberry Pis, and more. Installing Kubernetes on these machines and connecting them together creates a Kubernetes cluster. After creating a cluster, you can deploy applications to that cluster.

Machines in a Kubernetes cluster are, normally, referred to as Nodes.

Speaking of Nodes, a Kubernetes cluster contains two types of Nodes:

Master Nodes

Worker Nodes

Usually, Master Nodes are referred to as “Masters” and Worker Nodes are called “Nodes”.

Masters host the control plane and Nodes are where you run user applications.

The illustration below shows a 6-node Kubernetes cluster with 3 Masters and 3 Nodes. It is good practice for the Masters to exclusively run control plane services (no user applications). All user applications should run on Nodes.