Native AMQP 1.0
summary
RabbitMQ 4.0 embeds Native AMQP 1.0 in the core (no proxy plugin), yielding concrete wins: classic queues reach ~100k msgs/s (AMQP 0.9.1 ~10% lower), the old plugin performs ~4x worse, and the plugin needed ~7 GB extra for 40k connections. Link credit flow control (link flow control) enables per-queue backpressure so a consume benchmark moves from ~9k to ~663k msgs/s (≈70×). Built client wrappers exist for Java, .NET, Erlang and Go, and broker features use HTTP over AMQP and will add filter expressions in 4.1.
AI-written summary · about
chapters (19)
- 0:00Native AMQP 1.0 intro
- 3:14AMQP 0.9.1 to 1.0 evolution
- 8:55RabbitMQ adopting AMQP 1.0
- 11:43AMQP 1.0 in RabbitMQ 3.13
- 13:12AMQP 1.0 benchmarks
- 14:58Link flow control in 1.0
- 18:51Flow control 0.9.1 vs 1.0
- 23:52AMQP 1.0 flow control impact
- 24:53AMQP 1.0 flow control benefits
- 25:58Session flow control in 1.0
- 27:02RabbitMQ AMQP 1.0 clients
- 28:29HTTP over AMQP extension
- 31:00Sender settle mode outcomes
- 33:50Message immutability in 1.0
- 34:55AMQP 1.0 advantages summary
- 35:39AMQP 1.0 filter expressions
- 37:25Future AMQP 1.0 features
- 38:10Team contributions to 1.0
- 39:58Q&A filtering and clients
related talks
01h 28mHigh Performance Stream Processing
41mForget about HTTP
22mRack-AMQP: Ditch HTTP Inside SOA!
53mKeynote: AMQP Ten Years After
30mA love letter to messaging: Celebrating RabbitMQ's journey
31mWhat's New in Spring AMQP 2.0 - Gary Russell
59mAdding 1.21 Gigawatts to Applications with RabbitMQ
19mUpdate from RabbitMQ Core Team
