Scaling Kubernetes-based Event-driven Workloads with Keda & Karpenter
summary
Demo shows KEDA reading SQS queue depth and exposing metrics to HPA to spawn pods (the demo reaches ~58 scheduled pods with ~18 pending before nodes appear). Karpenter’s provisioner quickly launches matching capacity (spot G5 and C/R family examples) and can bin‑pack and replace interrupted spot instances. Uses EKS, k9s and an EKS node viewer; configs are a KEDA ScaledObject and a Karpenter Provisioner. Expect fast pod scaling but plan image startup and node sizing to avoid cold‑start delays.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:45Event-driven Stable Diffusion
- 5:58Live scaling demo
- 19:34KEDA and Karpenter together
- 27:04KEDA scaling patterns
- 30:36Karpenter for events
- 37:10Platform provisioning
- 38:57EKS blueprints
- 40:44Key takeaways and Q&A
related talks
38mScaling EDA Workloads with Kubernetes, KEDA & Karpenter
46mScaling EDA Workloads with Kubernetes, KEDA & Karpenter
29mCutting Costs and Boosting Efficiency with Karpenter and KEDA on Kubernetes
31mKubernetes workloads with Karpenter & Amazon EKS
46mKarpenter * Keda: The Winning FinOps Duo
22mKarpenter Efficient scaling of Kubernetes clusters
18mKarpenter : Your trusted Companion for Kubernetes Scaling
35mKEDA - Real Time And Serverless Scaling In Kubernetes