Pushing RabbitMQ to the limit
summary
Shows concrete throughput experiments using PerfTest on laptops and AWS EKS, with up to ~175k msgs/s on a fast Mac and ~150k combined with multiple queues, dropping to ~18k when heat throttling or cross-node consumers add network hops. Builds a PerfTest container, forces queue locality with a client-local master locator, and measures effects of publisher confirms (needs window ~100) and persistence/Quorum queues (single-node ~118k, Quorum drops to ~60k+). Finds large messages and slow disk drastically reduce throughput.
AI-written summary · about
chapters (5)
related talks
43mRealtime performance at scale (...)
33mReaching Ludicrous Speed
24mPushing Your Streaming Platform to the Limit
19mBreaking Storage Barriers: How RabbitMQ Streams Scale Beyond Local Disk
30mAbove the Clouds: Investigating Systems from 32,000 ft.
40mPerf Stories
01h 00mExploring Traffic Jams in Your Data Flows
43mRelease the locusts - distributed performance tests