If there were any doubt that containers are booming in the enterprise, a quick look at the tooling used to scale them should put that to rest. Apache Mesos, Docker Swarm, and Kubernetes are all growing fast, as enterprises seek out cluster managers to automate the deployment, scaling, and operation of containers.
Among these fast-growing open source communities, however, Kubernetes seems to be growing fastest.
The reason, suggests Apprenda executive Chris Gaun, is that Kubernetes offers a “very strong ecosystem that mimics the Hadoop model,” meaning that the “base tech [is] not sold as [an] installable solution by [the technology’s] originators but [instead is] adopted by vendors who productize it.” In other words, when there’s not just one dominant vendor on a project, the underlying project can flourish.
This thing has wings
Kubernetes, open sourced by Google in 2014, is based on Google’s Borg technology that has been running for many years. A decade before containers became trendy, Google had been powering its services with them.
SEE Hey ops teams, developers want control of the data center (TechRepublic)
Even so, Kubernetes was late to the container cluster manager party, with Apache Mesos first entering public discussion in 2009. Despite only being out of beta for 11 months, however, more people now list Kubernetes skills on their LinkedIn profiles than either Apache Mesos or Cloud Foundry. Interest can also be measured by the growth in Stack Overflow questions that mention the different container cluster managers, with Kubernetes dominating the category, as this chart from Apprenda shows:
The same trends hold true if we look at relative growth in jobs listing different container managers as a requirement.
The question is why?
Kubernetes and the community
It also helps that Kubernetes arguably provides the model by which Apache Mesos and Cloud Foundry developed. Cloud Foundry grew from Derek Collison’s experience working at Google with Borg. Apache Mesos, in turn, was arguably influenced by Ben Hindman’s experience working alongside ex-Googlers as they sought to re-create the Borg within Twitter. Everyone wanted to be Google.
SEE Mesophere tackles container orchestration in a big way (TechRepublic)
Now, however, Google has released the foundational code and invited others to productize it.
Perhaps the biggest reason for Kubernetes’ success is the fact that no one company owns it. Open source projects that gain the most community traction tend to be those that have a number of companies competing around them, as with Gaun’s Hadoop example, rather than a single entity contributing a disproportionate share of code.
Apache Mesos (Mesosphere), Cloud Foundry (Pivotal/VMware), and Docker Swarm (Docker, Inc.) each sees most of its development come from a single entity, something that no amount of foundation-washing can erase. This isn’t to suggest these projects aren’t doing well–they are, leading some like Gaun to posit that the container manager market won’t be winner-take-all so much as a market with a few commanding options.
Not everything is perfect in Kubernetes land, of course. Its documentation is poor, according to some users, and the shell scripts and installation options can be Byzantine. And yet, despite these problems, Kubernetes rolls on, backed by a decade’s worth of Google R&D and a community unshackled from heavy-handed vendor influence.