Magnum is an OpenStack service offering Container Clusters as a service,
with support for Docker Swarm, Kubernetes, Mesos or DC/OS. Users of the
service can deploy clusters of thousands of nodes in minutes, and access
them securely using their native APIs.
In the first part of this talk we introduce the Magnum service. We
present its basic and more advanced features, including support for
heterogeneous clusters, cluster upgrades, cluster drivers and the
ongoing effort to make it rock solid.
Next, you can get idea if how Magnum is integrated in CERN’s cloud
infrastructure. How it started as a pilot service, the efforts done to
add missing features and how CERN’s users benefit from it. CERN’s use
cases include end user physics analysis, web services, continuous
integration, infrastructure services and more.
Finally, such a complex service needs to be tested and benchmarked in
depth, so the last section includes the results of a series of tests
done in CERN cloud infrastructure and the CNCF testing infrastructure
scaling the different container orchestration engines to thousands of
nodes. Specifically, we report results of running several thousand
containers in Kubernetes and scaling a service to millions of requests a
second.