keda-gpu-scaler is a KEDA external scaler that reads GPU metrics from NVML on each node, so vLLM and Triton deployments scale on real GPU load instead of CPU, including scale to zero.
More: https://ku.bz/KB6slMgk4
Post #3988
683
- Sep 25, 2026This case study explains how Modelplane was built entirely as a Crossplane configuration,…
- Sep 25, 2026This case study shows how a team ran one Kafka cluster stretched across three separate Kub…
- Sep 24, 2026This case study shows how a payments company rebuilt deployments around GitOps with Argo C…
- Sep 24, 2026This case study shows how to migrate from fragile EC2-based infrastructure to AWS EKS with…
- Sep 24, 2026This case study shows how a nightly export job kept getting OOM-killed on Kubernetes, and…
- Sep 23, 2026This case study shows how one engineer built k8s-rightsizer in four days and cut Kubernete…