In episode 19 of our podcast, The Art of Modern Ops, we welcomed two renowned technologists to discuss the state of Kubernetes in the enterprise right now. 

Our first guest is industry figurehead and Kubernetes co-creator, Joe Beda. He is currently a Principal Engineer at VMWare where he’s been since their acquisition of Heptio, a leader in the cloud native movement, a company that he also co-founded. His career spans a variety of roles including engineer, founder, CEO and advisor. 

Our second guest is Michael Coté, a Staff Technologist at VMWare Tanzu (formally Pivotal). Michael’s background takes in stints as an analyst at 451 Research (now SPGlobal), Director of Cloud Strategy at Dell, and even a programmer where it all started.! He has written several books including the recent O’Reilly title, Monolithic Transformation and The Business Bottleneck.

Together, they offer different perspectives on the rise of Kubernetes in the conservative world of the enterprise. 

How the enterprise was won

On the podcast, Michael and Joe discuss the enterprise adoption of Kubernetes relative to its growth within the start-up and small-to-medium-sized business. Michael quoted a recent Gartner paper that referred to 10% of all enterprise applications now running on Kubernetes – proof that Kubernetes is a mature technology choice. The question they then ask themselves is why Kubernetes has been so successful? 

They agree that for all its complexity, Kubernetes is trying to make the infrastructure side of developing an application as simple as possible, by making most of the hard choices for you. It is opinionated to a certain extent, according to Beda, because it allows for other extensions of Kubernetes to flourish on top of it. 

“Kubernetes is being used in ways that we didn’t imagine when we first started the project. The drift towards edge, the use for ML... a lot of this stuff is a surprise to me ” – Joe Beda 

The power of community

Technical factors aside, they agreed that there has been a political dimension to Kubernetes’ success, too, because it’s been an open source project from very early on. This proved crucial because, in the early adopter space, the sense that there may be commercial interests at play could make potential users reluctant to invest their time and money in a technology. This is especially true for ambitious but cash-strapped start-ups, who are just beginning to build their platforms. 

“From the get go we saw this was a community project and we knew that the success was gonna be predicated on it running everywhere, on achieving ubiquity ” – Joe Beda 

In the enterprise, however, the adoption curve has been a little different. Michael and Joe suggest that this could be because, unlike the early adopter segment, which is populated by many thousands of start-ups and small businesses, only a very few enterprise level organisations – perhaps 50 – have been running Kubernetes at scale for more than five years. This has meant that a relatively small number of teams have been responsible for its adoption in the enterprise, and for communicating the success they have had. It is almost the opposite of the community-driven growth that has fuelled the rise of Kubernetes everywhere else. 

A compelling origin story

Ultimately, Joe and Michael concluded that regardless of segment, Kubernetes has been successful for three reasons. Firstly, because it did what it promised. It made it easier to build and manage cloud application infrastructure. Second is its open source nature. Finally, they considered an unintentional effect of its origins. While organizations of all sizes have contributed to Kubernetes’ growth, what many of them had in common – at some stage at least – is that they wanted to be Google. Kubernetes may not have had a commercial brand strategy, but its origin story, it could be argued, served as a brand. Here was a solution invented by the biggest and best in the world to solve their problems. For potential users, that meant it could also solve theirs. 

Looking ahead

“[Gitops is like] Here are a bunch of tools that will allow you to only make possible the enterprise architectures you want. There’s a lot more flexibility in what that enterprise architecture looks like. If you control the path to production, the build and how your containers and your configuration is deployed to production you basically are establishing an enterprise architecture because you limit what people can stick in that build pipeline. ” – Michael Cote

The conversation moved on to compare the declarative nature of Kubernetes and, more recently, GitOps, with the rigid, pre-Kubernetes ways in which applications used to be developed. And no conversation of this nature would be complete without looking forward, as well as back. So Michael and Joe spent a little time talking about the way they believe Kubernetes will evolve in the years to come – specifically in the context of how, via GitOps, it can be used to define and enforce an enterprise architecture across an entire organization. 

“If we look at Enterprise, one of the things that is just horrifying to me is that state of the art right now in many IT systems are these inventory systems where you can pay somebody and they’ll grovel around all of you machines trying to figure out what’s running on all of them so that you can inventory all the stuff that you are running. Now, with GitOps, with the declarative config we are turning that on its head. Instead of actually being inventory-driven – we react to what it’s running – we say, it shouldn’t be running unless it’s part of our declared config. This overtime will be fundamentally more secure. ” – Joe Beda

To hear the conversation in full, listen to the full podcast now: