KEDA

KEDA is a Kubernetes component that scales workloads from signals such as queue depth, request rate, or scheduled activity. It connects those signals to Kubernetes autoscaling so applications can respond to demand, including scaling eligible workloads to zero.

itCloud native tools and technologies

KEDA

Kubernetes can scale a workload from CPU or memory measurements through the Horizontal Pod Autoscaler, or HPA. Many workloads need a different signal. A message consumer cares about queue depth. A worker may care about pending jobs. An API may need a metric from Prometheus.

KEDA connects these event and external metric sources to Kubernetes autoscaling. You declare the workload, the trigger, and the scaling boundaries. KEDA polls the source, exposes external metrics to the HPA, and manages the HPA that changes the replica count.

This design keeps responsibilities separate. KEDA translates demand into metrics and handles activation between zero and one replica. The HPA handles scaling between one and many replicas. Kubernetes still schedules Pods and reconciles the workload.

The event-driven mental model

Think of KEDA as an adapter between demand and Kubernetes:

event source
    ↓
KEDA scaler
    ↓
external metric
    ↓
Kubernetes HPA
    ↓
workload replicas

A scaler understands one source type. Its trigger metadata defines how to reach that source and what value should cause scaling. KEDA includes scalers for queues, streams, databases, monitoring systems, cloud services, and other sources.

KEDA

The event-driven mental model

Relevant careers