First, there’s cross-cluster federated services — the ability to span workloads across clusters and, by extension, across multiple clouds.
It’s a logical move, given that the hybrid cloud — blending a private cloud with, most likely, multiple public clouds from the likes of Amazon Web Services (AWS) and Microsoft Azure — has become the default assumption for enterprise deployments. This opens up the possibility for workloads that need to draw resources from multiple clouds.
This would also mean that large jobs can be split among clouds. In fact, another feature of Kubernetes 1.3 is the ability to automatically scale services to match demand. The new version also doubles the maximum number of nodes per cluster.
Kubernetes also added support for rkt — the container runtime created by CoreOS as a Docker alternative — and for the Container Network Interface (CNI) and the still-in-process Open Container Initiative (OCI).
In other words, Kubernetes is being primed to support container frameworks beyond Docker. Nothing against Docker, but as usage patterns for containers evolve, alternatives are popping up — including not just rkt, but Windows containers or those based on the open source LXC project led by Canonical. It’s probably going to be a multi-container world for a while.
Other features inside Version 1.3 include additional resources for supporting stateful containers and a tool called Minikube that lets a developer fire up a Kubernetes cluster on a laptop.