Measuring Docker Adoption Rates Requires More Precision
How many people are adopting Docker for production use? That depends on how you define production. And that’s an important question for people tracing Docker adoption rates, which we could measure much more accurately than we currently are.
Over the past year, a series of surveys and reports (see here and here, for example) have predicted that container and microservice adoption rates are either skyrocketing, or on the verge of skyrocketing.
The latest example in this trend is a report out this week from 451 Research, which says the application container market will grow to $2.7 billion in 2020 from $762 million today.
But 451 Research also says container “adoption and revenue are just starting to emerge” and that “it is still early days for containers in the enterprise.” That means that the massive Docker adoption predicted by earlier reports has not come to pass, at least according to this analysis.
Given the fact that reports from early and mid-2016 predicted an imminent Docker explosion within the enterprise, yet 451 Research in early 2017 seems to think massive Docker adoption has not yet really started (but ostensibly will soon), you have to wonder: Why does one group of analysts keep making predictions about container adoption that the next group fails to see coming true?
Define Production
The answer, I think, has to do in part with the fact that these reports tend to define “production use” rather hazily. Docker adoption in the enterprise for production environments could mean lots of different things. It might mean just using containers for an internal app—which is less significant than running a business-critical, customer-facing app using Docker. Or an enterprise could report that it is using Docker “in production” by virtue of having moved just a handful of deployments to containers, while the vast majority of infrastructure remains old-generation.
Because of the ambiguity, it’s easy to see what you want about container adoption trends and ignore the rest.
To get a better sense of who in the enterprise is really using Docker for production, it would be helpful to view data on:
- Which percentage of a company’s apps are running on containers compared to those hosted using conventional technologies such as virtual machines or bare-metal. That would provide a more concrete idea of the extent of Docker adoption.
- How many business-critical apps are being run with Docker. Internal apps, or apps that are nice-to-have but not really essential, shouldn’t count toward Docker adoption rates.
- How many containers are involved in the applications that enterprises are deploying using Docker. A large, multi-container application deployment is more meaningful than a deployment in which a simple app is run inside a single container.
- Exactly which components enterprises are using to build their Docker stacks. This information would provide insight into which types of tools—orchestrators, registries, runtimes and so on—are most popular for Docker production use.
I do not mean to knock the reports that have been issued so far about Docker adoption. They’re useful. But they would be more useful if they eliminated some of the ambiguity regarding what Docker adoption actually looks like, and how Docker is being used in production.
