Communication in a microservice architecture
summary
Compares direct HTTP, RPC over a broker, and messaging with concrete benchmarks: same 20s test shows ~3,500 HTTP requests, ~5,000 RPC messages, and ~38,000 RabbitMQ messages, using AMQP and channels over one TCP connection. Explains at-least-once delivery, dead-letter queues, and RPC via queues, and shows RabbitMQ scales by adding consumers but shifts bottlenecks to the database (need for batching and connection-pool limits). Mentions actor models (Akka) and Kafka as alternatives and that messaging requires different data modeling.
AI-written summary · about
chapters (35)
- 0:00Introduction
- 1:39Monoliths
- 2:35Problems with monoliths
- 5:54Scaling
- 7:02Breaking boundaries
- 7:36Microservice vs monolith
- 10:40Communication
- 12:39HTTP calls
- 15:05HTTP calls between services
- 17:04Async calls
- 18:29HTTP
- 20:43HTTP Challenges
- 21:47RPC
- 22:03Message Broker
- 22:30Queue
- 22:48Queue demo
- 27:38RabbitMQ
- 28:53Async
- 30:21Database
- 30:45Batch processing
- 31:15Summary
- 32:40Scalability
- 33:22Performance
- 34:19Elasticity
- 34:56Flexibility
- 36:41Tools and frameworks
- 36:58Messaging middleware
- 37:33Queues
- 38:40Actor Model
- 40:11Message Brokers
- 42:36MPQ
- 45:03Direct exchange routing
- 45:55Fanout
- 47:07Enterprise Integration Patterns
- 48:20Messages
related talks
41mForget about HTTP
55mForget about HTTP
59mEffective Microservice Communication and Conversation Patterns
01h 03mChoosing the Right Messaging Solution for Your Architecture
27mApplication Integration Patterns for Microservices
50mMiddleware Showdown: Exploring Diverse Messaging Solutions
34mCall me maybe - communication between microservices
32mRabbitMQ in Real Life - Tips and Tricks