Microservices at Spotify
summary
Spotify runs about 810 active backend services powering roughly 75 million monthly active users, built so small teams can rewrite a service every few iterations. A Java-based framework called Apollo (now open sourced) plus a service registry called system z and a one‑button deploy tool nicknamed the provisioning cannon make creation, discovery and deployment fast. View aggregation services cut client latency, but thousands of tiny services raise monitoring, discovery and integration‑test challenges that the company is still addressing.
AI-written summary · about
chapters (22)
- 0:00Intro
- 0:48Thought experiment
- 2:08Product requirements
- 6:35Autonomous teams
- 11:31Scaling
- 14:23Microservices
- 17:44Failures
- 22:04Latency
- 24:36Scale
- 27:00Apollo
- 28:28Recap
- 31:07Questions
- 34:39Branching
- 35:18Integration
- 36:02Performance testing
- 36:53Automated tests
- 38:10Javabased services
- 41:24Architecture
- 42:28Migration
- 44:06Monitoring
- 45:45Code reviews
- 47:47Technical debt
related talks
46mModelling Microservices at Spotify
24mMicroservices in Practice
48mAgile at scale at Spotify
40mScaling code health with always on rewrites and AI at Spotify
27mBalancing autonomy, technical alignment, & speed through strategic collaboration
36mAdopting gRPC at Spotify
11mBreaking the (developer) silos
44mMicroservices: 5 things I wish I'd known