Killing the Dinosaur: A migration story
summary
Katch gets replaced by a new system called Skan that actually ships and runs in production, built to ingest Kafka into ECS + Java Spring, persist into DynamoDB + streams, enrich via AWS Lambda + SQS and serve searches from Elasticsearch. Migration uses AWS Glue to pull billions of rows from S3; the cutover runs in parallel and takes 1.5 years. Monthly ops drop from ~$25k to ~16–17k, but DynamoDB query limits, async consistency and Lambda sprawl remain real drawbacks.
AI-written summary · about
chapters (12)
- 0:00Start
- 2:09Warnings
- 4:42What was Katch?
- 8:34The Issues with Katch - The return
- 13:55Trust the process
- 15:19Tech Stack - Ingestion
- 18:39Tech Stack - Async Ingestion
- 21:22Tech Stack - Recap
- 22:04Existing Markets
- 25:24What's Glue?
- 30:09ElasticSearch vs DynamoDB
- 31:47AWS Lambda
related talks
31mWhen Bad Architectures Happen To Good People: Migrating Legacy Systems To Serverless
48mThe Journey of a strongly coupled Monolith into the Cloud Land
35mFrom Dinosaurs to Snakes - Decomposing the Enterprise Monolith
19mLearnings & Patterns from Zero-Downtime Migrations
41mTechnology mix: Around the serverless in 180 days
36m5 Hard-Learned Lessons from Successful Database Migration and Modernization
43mLegacy Architecture Migration Patterns
48mThe Great Migration