Currently this is more of a theoretical question than a practical problem. I am trying to figure out how extendable Kubernetes really is.

Think about something like an embedded device. This device has constraint resources and its firmware is difficult to upgrade. That's why it is not possible to run any Kubernetes logic on this device. The goal is to use this device, which communicates via IPv4 or IPv6, as a resource within Kubernetes.

A good example would be a k8s service that processes data from the embedded device on demand. But if there is no demand, it is not necessary to do anything with that data.

apiVersion: v1
kind: Pod
metadata:
  name: temperature
  namespace: temperature-example
spec:
  containers:
     - name: aggregate-temperature
       image: 
       resources:
           limits:
              temperature-sensor/v1: "400"
           requests:
              temperature-sensor/v1: "1"

I am aware of the device plugin but I am not sure if it fits my needs.

• Does it scale to <100k devices?

• Can it handle devices of the same type with different firmware versions?

• Is this useful for a lot of different device types (e.g. sensors for temperature, radiation, humidity, light, electricity, ...)

• Would it be possible to request a specific device e.g. by using its serial number?

• Could the cluster easily be extended by more device types?

If one of the questions is answered with no, what are the alternatives?