Reading the CNCF Landscape
The CNCF landscape is the best free map of cloud-native technology, and one of the easiest to misread as a shopping list. This six-part series shows how to read it capability first, what a CNCF maturity level does and does not tell you, how to score candidate projects, how managed services in South African cloud regions change the decision, the minimum platform around Kubernetes, and what is ready for AI workloads. It follows Baobab Retail Group, a fictional Johannesburg grocery retailer with a 15-person platform team; its systems and figures are illustrative. CloudNala's guidance is independent and not endorsed by the CNCF; project levels and regional availability were checked on 28 October 2026.
6 articles
- Part 1How to read the CNCF landscape without buying a logo wallThe CNCF landscape is the most useful map of cloud-native technology there is, and one of the easiest to misread. Part 1 explains what is actually on it, what a CNCF maturity level does and does not tell you, and a capability-first way to read it that ends in a shortlist instead of a logo wall.10 min read
- Part 2Graduated is a governance signal, not a support contractA CNCF Graduated badge tells you a lot about how a project is run and nothing about who will support your installation of it. Part 2 sets out what graduation actually evidences, the obligations that stay with you whatever the level, and how to write them down before an upgrade forces the question.7 min read
- Part 3A scorecard for choosing cloud-native projectsWhen two candidates share the same CNCF maturity level, the level cannot decide between them. Part 3 sets out an eight-criterion weighted scorecard, from maintainer diversity to local skills and exit cost, and works through two real-shaped decisions: a GitOps controller and a policy engine.7 min read
- Part 4Managed or self-hosted: cloud native in South African regionsAll three major hyperscalers now run managed Kubernetes in South Africa, but the services around it, the add-ons and the telemetry paths are not uniform. Part 4 sets out what we could confirm in each local region, how to think about residency without treating it as a checkbox, and where the real operational and rand-denominated costs sit.7 min read
- Part 5Kubernetes is the engine, not the platformA managed Kubernetes cluster runs containers. It does not route traffic sensibly, deploy from a reviewed source, enforce policy, show you what is happening or manage secrets. Part 5 sets out the minimum set of projects around the engine, their CNCF levels, a staged adoption plan for a 15-person team, and when not to use Kubernetes at all.6 min read
- Part 6Cloud native for AI workloads: what is ready and what to wait forAI is the fastest-moving corner of the cloud-native landscape, and the one where maturity labels need the most care. Part 6 checks the current state of the Kubernetes AI conformance programme, Kubeflow, KServe, llm-d, Ray and vLLM, shows how Baobab decides between managed AI services and self-hosting, and closes with how to watch the landscape for change.7 min read
Working through any of this in your own organisation?
These articles come out of delivery work rather than theory. If the problems in this series look familiar, a short scoping conversation is usually the quickest way to work out what is worth doing first.
Talk to CloudNala