DevOps at Uber
summary
Uber runs ~3,000 developers, ~2,000 microservices and 10,000 deployments per day across clusters with 100k+ containers; builds move from slow docker build to an internal build engine that cuts a 16m44s, 80GB image to half the time and ~60% of the size. Images distribute via a peer‑to‑peer system crackin that drops minutes to seconds. Cluster ops use Mesos with guaranteed + revocable tasks and rate‑limited scheduling to avoid cascading failovers. Metrics backend m3 ingests ~10M metrics/s; alerts are rules‑generated to reduce noisy on‑call pages.
AI-written summary · about
chapters (10)
- 0:00Introduction and background
- 0:31Uber as a global business
- 1:52The bits and atoms philosophy
- 3:42Scaling microservices
- 5:30Developer productivity tools
- 7:18Optimizing image builds
- 9:14Peer-to-peer distribution
- 11:18Efficient cluster management
- 14:20Monitoring and alerting
- 17:31Designing for scale and growth
