Knative

What it is

Knative is a Kubernetes-based platform for running serverless workloads without building your own scale-to-zero and eventing infrastructure from scratch. It has three parts: Knative Serving runs HTTP-triggered containers and autoscales them (including to zero) based on request load; Knative Eventing is a CloudEvents-over-HTTP routing layer for decoupling event producers and consumers; Knative Functions is a thin developer-facing layer on top of Serving/Eventing for writing functions instead of managing full container builds. It's not a PaaS and it's not FaaS-as-a-product — it's building blocks you deploy on your own cluster.

Who builds it and why

Originally a Google/Pivotal/IBM/Red Hat collaboration, now a CNCF graduated project with 336 contributors. Companies building internal platforms (PaaS teams, platform engineering orgs) contribute because they need scale-to-zero HTTP workloads and event routing on Kubernetes without reinventing autoscaling logic or writing their own CloudEvents plumbing. Red Hat's OpenShift Serverless is built on it, which explains a chunk of the sustained contribution. The motivation is straightforward: Kubernetes has no native concept of "scale a container to zero and back up on request," and Knative fills that gap in a vendor-neutral way.

Production readiness signal

CNCF graduated status is the strongest signal here — that requires demonstrated production adoption, a documented governance process, and security audits. 336 contributors and continuous commit activity (latest as of 2026-06-24) show this isn't a stalled project. Latest release is knative-v1.22.1, indicating a mature, actively versioned release cadence rather than pre-1.0 churn. The 6,066 GitHub stars figure is modest relative to its maturity level and adoption in commercial products (OpenShift Serverless, VMware Tanzu integrations), which suggests the star count undercounts real-world usage — a lot of Knative consumption happens through vendor distributions rather than direct upstream engagement.

Who should use this

• Platform teams building an internal developer platform on Kubernetes who need scale-to-zero HTTP services without writing custom autoscalers.
• Teams already committed to Kubernetes who want serverless UX (deploy a container, get a URL, get autoscaling) without leaving the K8s ecosystem or adopting a separate cloud FaaS product.
• Organizations building event-driven architectures that want a standard (CloudEvents) instead of bespoke pub/sub glue code.
• Teams running variable, bursty HTTP traffic where paying for idle replicas is wasteful.

Who should NOT use this

• Teams wanting a fully managed serverless experience with zero operational burden — you're still running and upgrading Kubernetes control plane components (Serving, Eventing, their CRDs, networking layer dependencies like Istio/Contour/Kourier). This is not "no ops."
• Latency-sensitive workloads that can't tolerate cold-start delays from scale-to-zero — Knative doesn't eliminate cold starts, it manages them.
• Small teams or single-app deployments where the operational overhead of running Knative's control plane outweighs the benefit — a simple Deployment + HPA is often enough.
• Anyone expecting AWS Lambda-level simplicity. Knative requires real Kubernetes expertise to operate reliably, especially around networking layer choice and eventing broker configuration.
• Organizations not already on Kubernetes — adopting Knative just to get serverless is adopting two platforms' worth of complexity at once.

Alternatives

• AWS Lambda / Google Cloud Run / Azure Functions — fully managed FaaS, no cluster to operate, but you give up portability and vendor neutrality.
• OpenFaaS — simpler FaaS framework on Kubernetes with a smaller footprint and less eventing sophistication than Knative Eventing.
• Kubernetes Deployment + HPA/KEDA — if you just need autoscaling (including scale-to-zero via KEDA) without the full Serving/Eventing abstraction, this is a lighter-weight path.

Pricing

Fully open source, Apache-2.0. No licensing cost. Your cost is the infrastructure you run it on (Kubernetes cluster, networking layer, eventing brokers) and the engineering time to operate and upgrade it. Managed distributions (e.g., Red Hat OpenShift Serverless) carry their own commercial pricing, but upstream Knative itself is free.