Practical advice for technical decision-making while evolving systems in place
summary
Fastly describes concrete outcomes: a metrics pipeline receiving thousands of emitters grows from 40 to 340 metrics, shifts aggregation from per-second to per-minute, and migrates storage from Bigtable to ClickHouse where validation catches an eight‑year race bug; architecture is rewritten 4–5 times, UI moves to a microfrontend architecture with a universal navigation and unified identity, testing and output validation prevent regressions, incremental rewrites succeed, and AI forces deeper investment in data interfaces and team competencies.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 1:25Evolve systems in place
- 3:30Metric pipeline principles
- 5:15Metrics pipeline growth
- 6:40Metric pipeline strategies
- 8:24Metrics pipeline architecture
- 9:29Control panel UI challenges
- 10:33Control panel complexity
- 11:58Monolithic UI to microservices
- 13:01UI navigation evolution
- 14:04Universal search expansion
- 14:45Identity systems evolution
- 15:47Microfrontend architecture
- 17:13System evolution summary
- 17:57AI as evolution driver
- 19:00Adapting to AI changes
- 20:21AI impact on data platforms
- 21:24AI in control panel
- 22:28Changeable vs sticky decisions
- 23:53Risk and security models
- 24:36AI adoption readiness
- 25:18Rapid change thoughts
- 26:42Conclusion
related talks
46mDigital scale-up architectural patterns
26mArchitecting for scale
30mEngineering in Overdrive: How to Thrive When Tech Evolution Goes Full Gas
49mBuilding Fastly's Platform for Scale: Continuous and Incremental
42mBetting On Evolutionary Architecture
31mEvolving performance assurance
26mBeyond the Hype: Facing Micro Frontend Challenges in Large Scale
52mBetting on Evolutionary Architecture