HPA

KServe Breaks Inference Autoscaling Into Three Paths as GPU Costs Face Scrutiny

KServe routes request-, resource-, and event-driven scaling to KPA, HPA, and KEDA, bringing GPU utilization and inference cost into focus.