This article is based on reporting originally published by devclass.com, citing the Cloud Native Computing Foundation (CNCF) annual user survey results published in early 2020.
Containers Are Now the Production Norm
The Cloud Native Computing Foundation’s annual user survey — covering 1,337 respondents drawn from the CNCF community — confirms what many practitioners already sense on the ground: containers and Kubernetes are no longer experimental technologies. They are mainstream production infrastructure.
Geographically, respondents skewed toward North America (38%), Europe (37%), and Asia (71% — note this figure likely represents a separate regional breakdown and may overlap with the other regions). The self-selected nature of the survey pool — CNCF community members — means results reflect an already cloud-native-leaning audience rather than the broader industry.
With that caveat in mind, the headline figure is striking: 84% of respondents were running containers in production, compared to just 23% in March 2016. That is a near four-fold increase in roughly four years. The only use case that actually declined was proof-of-concept work, dropping to 89% — suggesting that many teams that previously used containers experimentally have now moved them fully into production workloads.
Container Scale: How Many Are Teams Actually Running?
Scale varied significantly across respondents:
- 19% were running more than 5,000 containers in production
- 15% had between 1,000 and 4,999 containers in production
- 21% were running between 20 and 999 containers
On the platform side, Amazon EKS was the most widely used managed container platform at just under 30%, closely followed by Google Kubernetes Engine (GKE). The dominance of managed cloud Kubernetes services reflects the maturation of operational tooling and the preference for offloading control-plane management.
Kubernetes Adoption Continues Its Steep Climb
Kubernetes specifically saw strong year-over-year growth: 78% of respondents were using Kubernetes in production, up from 58% the prior year — a 20-percentage-point jump in 12 months.
Cluster counts give a sense of deployment complexity:
- 43% of Kubernetes users maintained between two and five clusters in production
- Just under 20% were operating a single cluster
Multi-cluster deployments — whether for availability zones, environment separation, or regional distribution — are clearly the dominant pattern for organisations that have moved beyond initial adoption.
Continuous Delivery: Faster Release Cadences Are Becoming Reality
One of the more compelling data points concerns release frequency — often the real measure of DevOps maturity:
- 27% of respondents were releasing daily, up from 15% the previous year
- 28% were releasing weekly, up from 20%
Automation adoption also shifted. Roughly 40% of developers were running fully automated releases (similar to the prior year), but there was a significant jump in teams combining automated and manual processes — rising from 25% to 41%. This hybrid approach often reflects organisations that have automated the pipeline but retained manual approval gates for production deployments.
On tooling, Jenkins remained the most popular CI/CD tool at 58%, followed by GitLab at 34% and CircleCI at 13%.
Cloud Strategy: Hybrid Enters the Picture
For the first time, the survey offered hybrid cloud as an explicit answer choice for cloud strategy. The result: 38% described their cloud as hybrid, while 62% opted for public cloud only. The introduction of this option alone makes year-over-year comparison difficult, but it confirms that a significant segment of cloud-native practitioners operate across mixed environments.
Serverless and Service Meshes: Growing, But Not Yet Universal
Serverless
Serverless adoption has grown to the point where non-use is now the minority position: only 34% of respondents reported not working with serverless technology. Meanwhile, 41% were actively using it, and 20% planned to adopt it within the next 12 to 18 months.
Service Meshes
Service meshes — which provide traffic management, observability, and security at the network layer between services — remained less mature in adoption:
- Only 18% of respondents reported actively using a service mesh
- 47% were in an evaluation phase
- The most popular options were Istio and Linkerd
The large evaluation cohort suggests service meshes are on many teams’ roadmaps but that operational complexity continues to slow broad adoption. The 47% evaluation figure also suggests the technology will likely see significant production uptake in subsequent survey cycles.
Context: KubeCon Amsterdam and COVID-19 Disruptions
The survey results were published against the backdrop of significant event disruptions in early 2020. The CNCF announced it was postponing KubeCon Amsterdam (originally scheduled for late March 2020) to July or August due to the COVID-19 outbreak. KubeCon Shanghai, scheduled for July, was cancelled outright. Major events from Google, Red Hat, IBM and others were similarly postponed or cancelled during this period.
Key Takeaways for DevOps Teams
- Containers in production are no longer a differentiator — they are table stakes for cloud-native organisations. If your team is still treating containers as experimental, the gap to peers is widening.
- Managed Kubernetes (EKS, GKE) dominates — self-managed control planes are increasingly rare among organisations that have hit scale.
- Release frequency is a real metric — teams shipping daily or weekly are no longer outliers. The infrastructure investment required to get there (automated pipelines, Kubernetes-native deployments, robust rollback) is increasingly well-understood.
- Jenkins still leads CI/CD, but the rise of GitLab suggests tight SCM/CI integration is a growing preference.
- Service meshes are the next wave — the high evaluation percentage suggests production adoption will accelerate, but teams should expect meaningful operational overhead when deploying Istio or similar tooling.
Survey data cited in this article is drawn from the CNCF’s annual user survey as reported by devclass.com in early 2020. For the primary source, refer to the CNCF reports page.