Moving 75,000 Engineers To DevOps On The Public Cloud
summary
Describes moving ~75,000 (now ~80,000) internal engineers onto a single One Engineering System (OneES/VSTS) that enforces branch+pull-request flow with ~80% PR pass rate, runs ~76,000 tests in ~16 minutes per PR, and collects ~5 petabytes/day of telemetry. Uses Azure across ~15 data centers with ringed progressive deploys and feature flags, enforces DRI on‑call rotations, and fixes scale issues by contributing GVFS for huge repos (≈300x git speedup). Emphasizes shift-left testing, production-as-definition-of-done, and team autonomy with enterprise alignment.
AI-written summary · about
chapters (11)
- 0:00Introduction and context
- 3:01Legacy engineering culture at Microsoft
- 6:17DevOps vision and principles
- 7:20Life of a change in OneES
- 19:22Scaling Git for large repos
- 21:31Open source contributions
- 22:39Cultural shift and team autonomy
- 25:31Public cloud and live site practices
- 27:17Feature crews and team structure
- 30:14Metrics and behavioral changes
- 31:42Outcome and culture transformation
related talks
07mWhat we learned moving 65,000 Microsofties to DevOps...
15mTransitioning 69,000 engineers to one cloud system
44m#Devops Journey
01h 02mEffective DevOps for Organizations
01h 01mAgile Evolution: An Enterprise transformation that shows that you can too
44mDevOps at Microsoft: Our Internal Transformation Story
06mGetting VSTS to "Production"
01h 02mBuild a DevOps Culture: Microsoft's Journey to adopt an Agile Mindset and DevOps culture