When Bad Architectures Happen To Good People: Migrating Legacy Systems To Serverless
summary
Argues a practical migration path: create an on‑prem agent to stream legacy DB changes into AWS EventBridge/Kinesis/SNS, consume with Lambda and build read models in DynamoDB while using DynamoDB streams to Kinesis for ETL into S3/Athena or Elasticsearch, and mirror writes back to MongoDB as a strangler pattern replacement. Recommends moving logs to S3, lifting stored procedures into Lambda, and using island‑hopping and a serverless ladder (lift, replatform, refactor). Cites a multinational POC that succeeded and spread adoption.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 2:22The Law of Server Lists
- 4:24Iceberg Diagram
- 5:13Greenfield Architecture
- 5:43Legacy Co
- 6:04Common problems to enterprises
- 6:59Unique technical challenges
- 7:49Human challenges
- 9:12Technical Challenges
- 9:33Event Sourcing
- 10:54New Systems
- 11:31Strangling Legacy APIs
- 13:38Migrating to the new version
- 14:20Detangling the database
- 15:05Access patterns
- 17:20Oversimplification
- 17:50Personas
- 20:36DevOps
- 21:13Traditional DevOps
- 22:06Serverless tribes
- 22:52Cloud lockin
- 24:28Douglas MacArthur
- 24:59The Less Defended Island
- 25:50The Fortification Behind You
- 26:26The Serverless Ladder
- 26:57Reposting
- 27:58How to form a new tribe
- 29:53The Builders Creed
related talks
45mPatterns & practices for developing enterprise-scale serverless applications
14mThinking Twice About Migrating to Serverless
23mSimplifying design complexities using Serverless Architecture on AWS
30mEvolving Serverless Architectures
49mServerless Architectural Patterns and Best Practices
16mFrom Complexity to Clarity
34mDesigning Data Intensive Applications In Serverless Architecture
42mEnterprise Serverless Adoption. An Experience Report