KCNA - Cloud Native Architecture (12% of the exam) - Section 4.3
Describe serverless and event-driven patterns in the cloud native ecosystem.
Functions as a service, scale-to-zero, event-driven autoscaling with KEDA, and Knative Serving and Eventing. Also when a serverless model is a poor fit.
serverlessKnativeKEDAscale to zeroevent-driven architectureCloudEvents
A team wants a framework that runs HTTP request-driven services, automatically scales their replicas up and down with traffic including down to zero when idle, and manages revision-based traffic splitting between versions of a service. Which cloud native project is purpose-built for this serving model?
AKEDA, an add-on that reads external event sources and drives autoscaling of existing workloads based on those signals.
BPrometheus, a monitoring system that collects time-series metrics and evaluates alerting rules against them.
CKnative, which runs request-driven services, scales them with traffic including to zero, and routes traffic across service revisions.check_circle Correct
DCloudEvents, a specification that defines a common metadata format so events can be described consistently across producers.
Identify Knative as the serving framework that runs request-driven services with autoscaling to zero and revision-based traffic routing. Knative Serving is designed to run stateless, request-driven workloads: it autoscales replicas based on incoming request demand, can scale a service to zero when there is no traffic, and models each deployment as a revision so traffic can be split or shifted between versions.
Why A is wrong: Tempting because KEDA also enables scale to zero, but it is an event-driven autoscaler layered onto existing workloads; it does not itself provide an HTTP serving framework with revision-based traffic routing.
Why B is wrong: Tempting because its metrics can feed autoscaling decisions, but Prometheus is an observability and alerting system and does not run application services or route request traffic between revisions.
Why C is correct: Correct: Knative Serving provides exactly this model, deploying stateless request-driven services, autoscaling them on request concurrency down to zero, and managing revisions with configurable traffic splitting.
Why D is wrong: Tempting in an event-driven context, but CloudEvents is a data-format specification for event metadata, not a runtime that hosts services or performs autoscaling and traffic routing.
Answers that look right on this material and are not. Each one is a distractor from a different question in the KCNA bank for this domain.
A standard Deployment with a fixed replica count of two, so a Pod is always ready to serve.
errorWhy it is wrong: It guarantees availability, which is tempting, but a fixed replica count keeps Pods running during idle periods and never scales to zero, so it fails the no-idle-resource requirement.
It is a serverless runtime that replaces the kubelet on each node so that functions execute directly without Pods.
errorWhy it is wrong: Tempting because it sounds like a serverless execution engine, but the described project does not replace the kubelet or bypass Pods; it drives the scaling of ordinary Kubernetes workloads.
Horizontal Pod Autoscaling, which adjusts replica count with load but keeps at least one replica running, avoiding any start-up delay.
errorWhy it is wrong: Tempting because the Horizontal Pod Autoscaler does add and remove replicas in response to load, but by default it holds a minimum of one replica and does not take a workload down to zero, so it is not the capability described.