What I Learned About Going Fast at eBay & Google
summary
Compares engineering practices at eBay and Google and gives concrete takeaways: App Engine runs ~3 million applications including ~15,000 internal apps; a hosted NoSQL stack (Cloud Datastore → Megastore → Bigtable → Colossus) lets a small team run massive services — one datastore team is ~6–8 people. Describes building a microservices chassis using Netflix OSS, Maven and AWS to spin a service up in ~15 minutes, and reports hundreds of parallel A/B tests that moved search ranking by tiny deltas (0.005%) to net a 2% revenue (~$120M) gain.
AI-written summary · about
chapters (32)
- 0:00Introduction
- 1:03How do I go fast
- 1:27Organizational culture
- 2:08User experience
- 2:28Less is more
- 3:26Quality and reliability
- 5:36Confidence in testing
- 6:42We dont have time
- 7:55Engineering at Google
- 9:28Bug reporting at Google
- 10:16Submissions from people Ive never met
- 11:09Service teams
- 12:46Service levels
- 13:43No sequel services
- 15:48Autonomy and accountability
- 17:05The chassis
- 20:09Open sourcing
- 20:53Team to Team
- 22:25Sit with the Team
- 23:30Process
- 24:57Postmortems
- 26:53Action Items
- 27:37Inversion and Experimentation
- 29:03Experimentation at eBay
- 32:32Technical tradeoffs
- 34:12People
- 37:25Hiring Process
- 40:27People arent interchangeable
- 42:12eBays development process
- 45:20Recap
- 45:37Questions
- 45:48Customer prioritization
related talks
49mWhat I Learned About Going Fast at Ebay & Google
48mMoving Fast at Scale
30mArchitecture & Teams at Less-than-Google Scale
18mMoving Fast at Scale
55mFollowing Google or Don't Follow the Followers, Follow the Leader
52mThe Future of Software Engineering
45mLessons from Android development at Google
30mBuilding Stable, Scalable & Agile Distributed Systems at Enterprise Scale