Kubernetes 1.2 makes large-scale Docker clusters 400 percent more scalable


Operating large-scale container clusters is set to become much more straightforward thanks to the new version of Google Inc.’s Kubernetes orchestration framework that hit the pages of GitHub this week. The release quadruples the maximum number of nodes in an implementation to 30,000 and brings a host of new automation capabilities designed to reduce the amount of work involved in maintaining so much infrastructure.

The most notable addition is a hands-off code deployment capability that enables Kubernetes 1.2 to manage almost every major aspect of updating an application by itself. Google says that the framework is able to register when a new version is available, implement the change according to custom specifications inputted beforehand and automatically undo the process if the installation fails for some reason. Multiple such rollouts may be performed at the same time, which removes the need for a developer to wait on a colleague’s patch to be applied before theirs is released.

Updating application settings can be done faster as well thanks to the fact that Kubernetes now no longer requires recomposing every underlying container to perform a change. On the downside, the capability might make it easier for configuration mistakes to slip through, but the new failover functionality introduced in conjunction should help mitigate the risk. If a Docker instance becomes online due to a bad patch or some other technical issue, the framework is able to fire up another container in an entirely different region. The functionality should come particularly handy in off-premise clusters running on a cloud platform like AWS that occasionally suffers localized data center outages.

Google hopes that the new features will help Kubernetes compete more effectively against Docker Swarm, which was shown to be vastly faster at spinning up containers in a recent benchmark study sponsored by its creator. But while the search giant’s focus on ease-of-use may win over some developers, it’ll need to catch up with its rival on performance as well if the project is to maintain relevance on the long run. The company appears to be well aware of this fact judging by the roadmap for the next version of Kubernetes that was unveiled alongside the 1.2 release.

In addition to speed improvements, the next iteration of the framework is set to bring enhanced support for stateful applications like databases that have to retain information for long periods of time. Google is also promising to introduce a capability for coordinating disparate container clusters that will make it more feasible to use Docker in hybrid cloud use cases.

Photo via Andi Graf