Building a platform to integrate serverless and on-premise services
summary
A platform called Scenario integrates cloud functions and legacy on‑prem jobs by cataloging services, providing a graphic workflow editor, and running an execution engine with task queues and a cockpit for live state and inputs/outputs. It supports long‑running jobs (hours→weeks) with partial restart of subgraphs, deploys via Helm to Kubernetes, and uses codeful connectors to call AWS Lambda, Azure, OpenFaaS or internal batch systems while the platform keeps credentials local with connectors keep secrets local. JavaScript sandbox tasks transform outputs between nodes.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:14Microservices products
- 2:20Customer constraints
- 5:58Long-running tasks
- 9:15Integration challenges
- 13:37Why external tools fail
- 17:08Dedicated integration platform
- 19:37Platform design
- 21:28Execution engine challenges
- 24:19Connectors for services
- 27:50Live demo
- 34:06Building scenarios
- 42:24Key takeaways
- 43:30Final demo
related talks
27mBuilding & Running a Serverless Platform: Beyond Infrastructure
39mDevelopment productivity in a post serverless world
39mBuilding Serverless Application Pipelines
29mBuilding a Serverless Compute Platform with AWS and TypeScript
35mMulti-Cloud serverless deployment
25mWe accidentally created a Serverless Application
01h 02mServerless Architecture Patterns
41mBuilding and Running an Enterprise-grade Serverless Platform on Kubernetes