Taking the Mystery Out of Container Terminology, from Kubernetes to FaaS

Written by Andrew Gracey
November 14, 2019

A guide to the lingo: C is for container, K is for Kubernetes.

… smaller bits of code to events and expose those events in a variety of triggers (for example, HTTP or publish-subscribe messaging, also called PubSub). This allows you to scale different pieces independently, as well as scale to zero between events to save on compute costs. Done right, it can be a very effective way to process data.

As with a service mesh, a FaaS or serverless framework abstracts a lot of repeated components pieces into a single layer.

OCI, CRI and Container Runtimes. Fairly early on, the Kubernetes team realized that it was a bad idea to keep implementation details about how containers are run in their code itself. This meant creating some specifications around how to communicate with potential runtimes. Out of this, we got three specs:

Container Runtime Initiative (CRI), allowing control from Kubernetes-compliant orchestration.
Open Container Initiative (OCI) Runtime Spec, governing how to communicate with a runtime.
Open Container Initiative Image Spec, giving details on how to build, store and transfer container images.

Networking and Storage. Similar to the runtime, handling networking and storage is done by the Container Networking Interface (CNI) and Container Storage Interface (CSI), each exposing their own respective services.

Each of these have API objects that allow common administration across providers to give your users easy configuration. That said, some providers will allow extra configuration through Custom Resource Definitions (CRD).

Next Steps

Now that you better understand these terms that get thrown about, where do you start?

The best place to start might be to spin up a small cluster and see how easy it can be to install software into it. There are a variety of good distributions out there fitting a variety of needs that will ease the installation and configuration pain.

Once you have a cluster, see if there’s a helm chart for a piece of your existing infrastructure. Spin it up, see how it works and maybe even pull some traffic on to it once you are comfortable.

Andrew Gracey is a technical marketing manager at SUSE with 10-plus years of experience in a variety of software engineering positions. He is interested in the intersection of business, technology and human interactions. Follow Andrew on LinkedIn or on Twitter @gracey_andrew or @SUSE.