managed clusters #33
Description
When dealing with a consolidated distributed computation framework, such as Spark, we could deploy it standalone, as the cluster manager itself.
However, there are also options of "managers" as backends, like Yarn, Mesos or Kubernetes, where they serve as resource providers to Spark.
From what I could understand, Constellation aims to be able to be deployed directly to a group of machines connected by network.
Would it also work on cluster managers? One great advantage of this would be the elasticity feature from cloud providers, or healthchecks/healing from the cluster manager itself.
I understand targeting bare metal could harness more performance by escaping management overhead, but the practicality of having it run on a managed environment could be a nice option to some of us.
Anyway, congrats on the brilliant work!