Scaling EDA Workloads with Kubernetes, KEDA & Karpenter
summary
Explains a concrete pipeline that uses EKS + KEDA to scale pods from zero based on queue depth (SQS/Kinesis) and Karpenter to add right‑sized EC2 capacity, bin‑packing unschedulable pods and choosing spot or on‑demand for cost. Shows numbers: ~3.5M pre‑registrations and real incidents with ~4× expected peak traffic, config knobs like polling interval and scheduled pre‑scale, and a working demo where KEDA triggers HPA and Karpenter provisions nodes so queued payments complete. KEDA scales on queue length Karpenter bin‑packs CPU needs
AI-written summary · about
chapters (9)
- 0:00Intro
- 2:14Agenda
- 2:42The Taylor Swift problem
- 16:16Scaling with Kubernetes
- 18:30KEDA overview
- 26:17Application scaling with Karpenter
- 32:47Solving our Taylor Swift problem
- 36:48Key takeaways
- 37:48Outro
related talks
42mScaling Kubernetes-based Event-driven Workloads with Keda & Karpenter
29mCutting Costs and Boosting Efficiency with Karpenter and KEDA on Kubernetes
46mKarpenter * Keda: The Winning FinOps Duo
37mKEDA: Unlocking Advanced Event-Driven Scaling for Kubernetes
40mOptimize EKS Cost and Scaling with Karpenter
37mKubernetes Event-driven Autoscaling with KEDA
36mAutoscaling Isn't Set-and-Forget: Tuning KEDA & Karpenter for Real Workloads.
25mHow to take your Kubernetes cluster to next level with Karpenter, Ray and Amazon EKS?