Didn’t do this

Thanks to Florient Chouteau and Dennis Wilson

for their work on this subject.

I took most of the content from theirs:

• https://supaerodatascience.github.io/DE/slides/1_5_orchestration.html
• https://supaerodatascience.github.io/DE/slides/2_3_kubernetes.html#/

Why orchestration?

Suppose I have a large pool of machines available

• How do I deploy my container?
• How do I put the right containers at the right spot?
• How do I scale (up and down) to demand?
• How do I expose the http endpoints?
• How do I manage failure of containers?
• How do I update my model without downtime?

Tools example

Examples…

• Docker Swarm
• CoreOS Fleet
• Apache Mesos / Marathon

… and so many more !

Short Kubernetes history

Kubernetes (or k8s) comes from Google’s internal systems Borg

It is open source now https://github.com/kubernetes and used… everywhere ?

Thanks to Dennis for the next part

Slides

Quizz

What allows container orchestration (multiple choices)?

• Answer A: Auto scaling of services
• Answer B: Making coffee
• Answer C: Ensuring container states and availability
• Answer D: Submitting finite jobs

Answer link Key: vp

Kubeflow

https://www.kubeflow.org/

Jupyterhub/Binder

Databricks: Spark as a Service

Coiled: Dask as a Service

Quizz

How do you deploy containers through Kubernetes (mutliple choices)?

• Answer A: docker run
• Answer B: Helm charts and command
• Answer C: yaml configuration and kubectl
• Answer D: custom bash script

Answer link Key: gb

Try some

https://training.play-with-kubernetes.com/kubernetes-workshop/

https://labs.play-with-k8s.com/ with Voting App