Serverless on Kubernetes: What We Learned From Customers
summary
Azure describes how customers run serverless on Kubernetes using the func tooling to auto-generate container images and KEDA to replace the cloud scale controller so code scales 0→n via HPA and event scalers; durable functions orchestrate parallel ML tasks for Fujifilm and reduce photo-availability from hours to <2 minutes; a Canadian broadcaster hits 1 million req/s on HTTP flows; KEDA ships with a dozen+ event sources and proves portable serverless works on-prem and multi-cloud.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 1:54Our journey
- 2:46HTTP serving
- 4:17Canadian election
- 5:14Multilanguage
- 6:01HTTP in
- 6:30Eventdriven applications
- 7:05Web jobs
- 8:55Eventdriven
- 10:05Functions
- 13:55Scaling
- 14:52Eventdriven scale
- 15:08Fujifilm
- 16:23Durable Functions
- 20:08Kubernetes
- 21:30Interviewing Application Developers
- 22:06Kubernetes Ecosystem
- 24:12Spring
- 25:08Combining Technologies
- 26:58Call for Contributions
related talks
39mServerless on Kubernetes with Kubeless [A]
57mServerless on AKS (Azure Kubernetes Service)
41mIs Serverless the End of Kubernetes?
52mServerless landscape beyond functions
01h 00mServerless Madness on Kubernetes
21mKeynote: Kubernetes Serverless Present and Serverless Future
29mGoing Serverless
30mUnleashing the Power of Serverless on K8s with Knative, Crossplane, Dapr, KEDA