The (big) picture of debt
summary
Case study of splitting a shared Postgres instance into a Postgres as a service platform: claim is that you must first fix platform and developer ergonomics, not just yank schemas. Built an API framework (18 months to mature), automated most migration changes, and executed a phased plan—draw ownership boundaries, then reduce access and migrate teams—over almost a year while the company grows from ~2,000 to ~5,000 engineers. Result: service goes GA and migrations proceed incrementally with narrow blast radii.
AI-written summary · about
chapters (10)
- 0:00Technical debt basics
- 3:12Shared database pet peeve
- 7:26Avoiding new DB instances
- 10:31Monolithic DB problem
- 13:19Datadog storage frameworks
- 16:31DB migration strategy
- 20:50Incremental migration plan
- 23:39Project handoff rewards
- 24:42Lessons from debt tackling
- 26:29Closing thoughts
related talks
52mFrom 0 $100 billion: Inside lessons on how to scale your dev workflow
32mBuilding Platform Engineering from scratch at scale up
50mScaling for Global Growth with Modern Cloud
13mManaging the marathon: Leading teams through migrations
40mPlatform Wars: The Battle Between Golden Paths and Spaghetti Pipelines
32mStarting and Scaling a Platform Engineering Team
45mLeading a High-Performing Postgres Team: Culture of Excellence & Growth
30mProduction › Prototype: How a Small Team Replaced a Core Platform in 90 Days