Micro Services and Events: Friends or Foes?
summary
A shared‑nothing microservices platform actually runs about 60 Elixir microservices and uses RabbitMQ + Broadway for event delivery, protobuf schemas and an S3-backed archive as the event ledger (protobuf schemas + S3 event store). It processes bursts up to 10k events/sec (typical 1–2k/sec), enforces at‑least‑once with daily 24h replay to heal duplicates and out‑of‑order deliveries, uses upserts with microsecond timestamps, retired delta updates, and achieves five‑nine uptime while scaling BI via a streamed data lake.
AI-written summary · about
chapters (7)
- 0:0006:51 - Meaningful Conversations at Scale
- 6:5209:56 - Why Microservices?
- 9:5713:42 - Why Event-Based and Event-Sourced?
- 13:4318:39 - Technology & Architecture
- 18:4033:04 - Lessons Learned
- 33:0535:40 - Next Steps
- 35:4144:38 - QnA
related talks
27mScaling a Rails Monolith w/Event Driven Microservices
16mMicroservice Challenges
03mMicroServices - Let’s Build Some!
24mMicroservices in Practice
49mChallenges in Implementing Microservices
15mMicroservices with Elixir
48mChallenges in Implementing MicroServices
01h 03mMicroservices vs The Distributed Monolith