Which autoscaling solution can scale workloads down to zero pods in Kubernetes?

Which autoscaling solution can scale workloads down to zero pods in Kubernetes?

• A. Kubernetes Event-Driven Autoscaling (KEDA)
• B. Horizontal Pod Autoscaler
• C. Vertical Pod Autoscaler
• D. Cluster Autoscaler

Answer: (A)

Scaling to zero pods happens when there is no work to do, and the solution that handles this through external event-driven triggers is Kubernetes Event-Driven Autoscaling (KEDA). KEDA watches external sources like message queues, Kafka topics, or HTTP events and uses a ScaledObject to adjust a deployment's replica count based on those events. When the queue is empty or no events are arriving, it scales down to zero pods, and it scales back up as soon as new work comes in. This makes it ideal for workloads that are intermittent or driven by outside events. The Horizontal Pod Autoscaler, on the other hand, scales pods based on metrics from within the cluster (like CPU or latency). It can reduce replicas, but typically requires a nonzero minimum unless you explicitly set minReplicas to zero, and it isn’t inherently event-driven. The Vertical Pod Autoscaler changes resource requests inside pods rather than the number of pods, so it doesn’t reduce pod counts to zero. The Cluster Autoscaler changes the number of nodes in the cluster, not the number of pods running on those nodes.

Scaling to zero pods happens when there is no work to do, and the solution that handles this through external event-driven triggers is Kubernetes Event-Driven Autoscaling (KEDA). KEDA watches external sources like message queues, Kafka topics, or HTTP events and uses a ScaledObject to adjust a deployment's replica count based on those events. When the queue is empty or no events are arriving, it scales down to zero pods, and it scales back up as soon as new work comes in. This makes it ideal for workloads that are intermittent or driven by outside events.

The Horizontal Pod Autoscaler, on the other hand, scales pods based on metrics from within the cluster (like CPU or latency). It can reduce replicas, but typically requires a nonzero minimum unless you explicitly set minReplicas to zero, and it isn’t inherently event-driven. The Vertical Pod Autoscaler changes resource requests inside pods rather than the number of pods, so it doesn’t reduce pod counts to zero. The Cluster Autoscaler changes the number of nodes in the cluster, not the number of pods running on those nodes.