Kubernetes Crosses the Chasm, and Other Lessons from the 2021 CNCF Survey

You know I love a good survey, so let’s take a look at the Cloud Native Computing Foundation’s 2021 annual survey. They asked 2,302 respondents how they use Kubernetes and the more general category of cloud-native tools. The major conclusion of the report is that Kubernetes usage is mainstream, as the sub-title of the report labeled 2021: “The year Kubernetes crossed the chasm.”

This year’s survey has numerous topics, but let’s look at the three topics that interest me most.

Going mainstream

When the CNCF’s survey sports the strapline “the year Kubernetes crossed the chasm,” what this means is “Kubernetes is entering the mainstream.” Let’s check that claim. First: “chasm”?

Related to the diffusion of innovations theory, Geoffrey Moore’s Crossing the Chasm work posits that software market adoption goes through five phases: innovations, early adopters, the early majority, the late majority, and the laggards. Each group is progressively less tolerant of new, harder-to-use software. Innovators will put up with anything because they like using new technologies, laggards may not even be interested in your new software. There’s a chasm between visionaries and the early majority that most software companies can’t cross. This is where what worked for a vendor (now, public cloud as well) with that small set of customers no longer works as well for everyone else. They have to “scale” their product management, engineering, support, sales, and marketing to expand into organizations that have increasingly everyday, normal, indeed, boring concerns. Moore’s model, then, is a good one for figuring out if and when Kubernetes has gone mainstream.

Let’s do the math then to see if Kubernetes has made it across the chasm. Each group in Moore’s model shows up (surprise!) in a bell curve with innovators representing 2.5%, early adopters 13.5%, the early majority 34%, late majority 34%, and laggards (thankfully!) just 16%. To have crossed the chasm, then, you’d need 16% of people using your software. If you were in the range of 17% to 50%, you’d be into mainstream territory.

Now, let’s calculate how many global developers are using Kubernetes. SlashData, who conducted this survey for the CNCF, estimated that there were 26.8 million developers globally in 2021. The CNCF survey says there are “5.6 million developers using Kubernetes today.” Thus, we get the following pie chart:

If my math is right, 21% is greater than 16%, so this gets Kubernetes across the chasm. Now, a word of caution: I don’t know how SlashData and the CNCF came up with that 5.6 million developers using Kubernetes estimate. But let’s keep going.

I think we all know intuitively that Kubernetes is at the beginning of its mainstream life, and it’s always good to have some numbers like these to back that up.

However, for something like Kubernetes, I like to keep an eye on the share of all applications: out of all the applications running worldwide, how many are running on Kubernetes? So far, this pie chart has eluded me. Some analysts have taken a go at it, but I don’t think we have a solid grasp on Kubernete’s share of all applications yet. Or maybe the numbers exist and I just need to renew some of my analyst seats and pore over their tasty pies. Hopefully, the next survey will tackle the question of application share.

2021 is the year of serverless…again

With 39% of respondents doing serverless, this is clearly the year of serverless…but that's been the proclamation for the past 5+ years. Sarcasm aside, while the growth is small, it is clearly increasing. Serverless is definitely “a thing,” but it’s been at the same general usage rate for several years now.

Since 2016, if not earlier, each year serverless has always been on the verge of changing everything. Is 2022 going to be the year of serverless? In the 2021 survey, 39% of respondents said they’re using serverless technology. Let’s see how this compares to previous years, in the chart below.

Clearly, something weird happened in 2020, but the overall trend is level. Little change in usage over the years could make you think that all the people who are going to use serverless are already using it. Adding to that, O’Reilly reported a steep decline in serverless training last year. Or, you could be more optimistic: if you look at the evaluating and planning answers over time, you might theorize that there’s a second wave of people who are interested in serverless but aren’t using it yet. Carving out that story more, you could say this cluster is the late majority and the laggards. However, because there’s such variation in these numbers, I wouldn’t be too precise in my conclusions. We’ll have to wait and see what the next few years have in store.

While we wait, we can look at the public cloud serverless options that people use for serverless. Amazon, Azure, and Google top the list for hosted serverless services. This has been the same in the surveys since 2018. Looking at “installable software” for serverless over time is much more fun because you see lots of coming and going (put that lid on your margarita for this one):

In this chart, I only included the top three each year (except 2021), carrying them over to the following years to see how they move. This left many options off, for example, in 2021 there were 15 frameworks to choose from, including “other.”

What you see over the years in serverless framework use is that OpenFaaS has stuck it out and that Knative has quickly risen in popularity. However, I wouldn’t start breaking champagne bottles yet: in the 2021 survey, only 151 people answered this framework question with 1,376 people skipping the question. There are a lot of people yet to consider serverless.

Time-to-Market Check: Software Release Cycle

I’m always interested in how long it takes organizations to go from idea to software in production. The more frequently you can release software, the more feedback you’ll get about the usefulness of the software because you can observe people using it more and adjust the features accordingly. There may be applications that are exceptions, but I haven’t seen them yet. How is the community doing with software releases? 

While it’s dangerous to draw causation between using Kubernetes and software release cycles, the people who responded to this survey are doing well when it comes to frequent deploys. In this year’s survey, over 80% of respondents say they release code to production at least monthly, with 31% doing it weekly and 33% daily.

While gradual, there’s a great trend of shortening release cycles over these four years. This is key for improving the design and effectiveness of your software because you can put a small batch loop in place that allows you to experiment with new features each week learning what works and doesn’t so you’re always making your software better.

Wrapping Up: This is Not a Zero Sum Game

After watching the recent Kubernetes documentary—which is great, you should check it out—it’s worth pondering the result that Amazon Web Services tops the list of public cloud Kubernetes services and some of the other questions and responses. In the Kubernetes documentary, among many other motivations, you hear that Kubernetes was initially a competitive response to Amazon. However, Amazon not only eventually joined the Kubernetes party, but has become the top used distribution and hosted Kubernetes service. This echos a major point of the documentary, that Kubernetes is not a zero-sum game: every interest “wins,” especially the users.

There’s a lot more in this year’s survey, and certainly more in the raw data if you’re into that kind of thing. And if you are into that kind of thing, also check out my write-up of our 2021 State of Kubernetes survey as well.