Abstract:

Not everything which can be build in containers is a good idea to be shipped to production. It might make sense to test a Hadoop pipeline with data in dev and test environment, however is this a good idea to ship this to production?

Containers are different from virtualisation, and their life cyle should be treated differently.

Should containers be run on our current package based Linux distributions or is the paradigm shift so radical, that we should even rethink the way, we run our platforms. Is the additional effort of container orchestration with kubernetes, etcd, flannel,... another layer of complexity to be fought with or is it a strategy to simplify our distributed systems in the long run?

If we want to go this way, what are the next steps in operations to go forward on this road?

The talk discusses patterns and antipatterns of container rollouts. With some two years of experience with containers in general and especially Docker it is time to review container rollouts from a DevOps point of view. The talk tries to raise awareness, that containerized systems need also a new architecture for the data center of the 21st century.

Speaker: Thomas Fricke

blog comments powered by Disqus