Tools for simplified large scale systems
summary
Argues concrete levers for scaling: adopt a single discovery catalog like Backstage (or an in‑house Revolt Tower) with SLO/SLA, automate updates via CI/CD, and enforce a constrained stack (JVM/Java, Postgres, Kubernetes) to reduce randomness. Shows a real refactor: a 3k‑endpoint monolith with 500 consumers and 20 teams cut build/feedback from over an hour to 10–20 minutes. Recommends blameless postmortems, RACI/ownership rotation, and focusing refactors on the single biggest bottleneck.
AI-written summary · about
chapters (17)
- 0:00Introduction to culture engineering
- 1:52Westrum organizations explained
- 5:46Speaker introduction and background
- 7:13Revolut's hyper growth context
- 10:27Problem of knowledge discovery
- 12:11Service discoverability tools
- 15:40Systems thinking in software
- 19:59Anti-patterns of technology diversity
- 23:09Simplified architecture at Revolut
- 27:04Signs of blurred ownership
- 33:26Practical tips for better ownership
- 37:04Incident postmortems and culture
- 40:35Dependency hygiene at scale
- 45:49When to refactor architecture
- 51:10Case study: Monolith refactoring
- 52:56Key takeaways for high performance
- 56:27Q&A session
