Building distributed system for retail with RabbitMQ
summary
Real retail system runs in production for three years using RabbitMQ with a cloud cluster and around five-node cloud cluster, roughly 700–900 on-prem edges across environments. Messages are small (up to ~800 bytes) at about 50 msg/s average and measured ~536 ms latency for edge round trips. Initial federation fanout links cause connection storms so design moves to per-edge direct exchanges plus shovel, randomized reconnect backoff, durable queues and TTLs, which cuts connections and stabilizes the rollout.
AI-written summary · about
chapters (6)
- 0:0005:35 - Introduction
- 5:3608:01 - What are the use cases?
- 8:0213:48 - What is under the hood?
- 13:4718:01 - System in numbers
- 18:0223:06 - Cloud cluster
- 23:0731:59 - QnA
related talks
19mBuilding a Distributed Data Ingestion System with RabbitMQ
44mManaging RabbitMQ Usage at Scale for Bloomberg’s Trading Systems
24mAdvanced RabbitMQ Strategies: A case study
32mRabbitMQ in Real Life - Tips and Tricks
48mBuilding a Distributed Data Ingestion System with RabbitMQ
22mLessons Learned from Scaling RabbitMQ in a High-Growth (...)
40mScaling RabbitMQ
48mA Beginner's Guide to Inter-Service Messaging