Event-driven autoscaling for Serverless Java

on YouTube ↗49m

" summary

Kubernetes HPA based on CPU proves too slow for bursts: a Quarkus Java consumer faces 5 million Kafka events and HPA lags while KEDA scrapes Kafka/event metrics and triggers HPA to jump to 10 pods quickly; Knative provides true scale-to-zero but only starts at one pod unless combined; demo wiring uses Quarkus + Kafka + KEDA + Knative, shows KEDA handling high fan-out and Knative delivering cold-start-free cost savings when reworked, and the KEDA+Knative integration gives flexible, working event-driven autoscaling.

AI-written summary · about

" chapters (9)

1. 0:00Introduction
2. 2:20Kubernetes limits
3. 6:55Introducing KEDA
4. 13:45Demo: HPA basics
5. 23:23Optimizing with KEDA
6. 27:57Demo: Knative serverless
7. 39:53KEDA + Knative
8. 47:01Choosing autoscaling
9. 48:48Q&A

" related talks

• 56mEvent-driven autoscaling for Serverless JavaDaniel Oh · GeeCON 2023
• 58mEvent driven Autoscaling for Serverless JavaDaniel Oh · Great International Developer Summit 2023
• 52mEvent-Driven Autoscaling for Serverless JavaDaniel Oh · JCON Europe 2023
• 35mKEDA - Real Time And Serverless Scaling In KubernetesZbynek Roubalik, Jeff Hollan · KubeCon + CloudNativeCon North America 2022
• 37mKubernetes Event-driven Autoscaling with KEDAZbynek Roubalik, Jorge Turrado · KubeCon + CloudNativeCon Europe 2022
• 36mKEDA: Event Driven and Serverless Containers in KubernetesJeff Hollan · KubeCon + CloudNativeCon North America 2019
• 34mAutoscaling Made Easy with Kubernetes Event-Driven Autoscaling (KEDA)Tom Kerkhove · Global Azure Virtual 2020
• 46mScaling a dormant Java application? Quarkus and Knative to the rescue!Kevin Dubois · Voxxed Days Luxembourg 2022