What it Takes to Build & Operate a Global Serverless Provider
summary
Platform abstracts public clouds and orchestrates workloads with Nomad orchestration and lightweight VMs via Firecracker microVMs, running containers inside microVMs with an Envoy sidecar and Kuma service mesh for mTLS, native L7 load balancing and DNS. Control plane issues jobs to Nomad; data plane runs customer workloads. Global footprint is 250 edge locations with six core regions and the UI/API deploy six regions in ~4 minutes; TLS, HTTP/3, automatic domain and private networking are provided.
AI-written summary · about
chapters (11)
- 0:00Introduction and overview
- 3:13What our platform does
- 5:01Why global deployment matters
- 6:30Demystifying serverless
- 8:19How the platform works
- 10:06Key technology choices
- 13:00Under the hood architecture
- 16:17Service mesh and networking
- 17:22Global deployment demo
- 21:39How traffic routing works
- 23:04Q&A and closing
related talks
01h 01mTaming the Modern Public and Private Clouds with Nomad
20mComposing the Cloud, 2.0
33mHow to Recreate AWS in Your Data Center
48mScaling for Global Growth with Modern Cloud
35mMulti-Cloud serverless deployment
13mSam Kroonenburg, A Cloud Guru | ServerlessConf 2018
39mServerless Platform for Large Scale Mini-Apps: From Knative to Production
35mWhat does Serverless mean for tomorrow’s abstracted infrastructure?