A Murder of LiveViews: Distributed Load Testing with AMoC and FLAME

A little over a decade ago, Phoenix proved that it could handle 2 million WebSocket connections. A lot has changed since then, and with LiveView, the paradigm has shifted! Connections aren’t the only bottleneck anymore, your LiveView architecture is. How you structure Functional Components, LiveComponents, Streams, and Async Assigns all affect performance under load. How do you know your app can handle real traffic? This talk introduces a distributed approach using AMoC (“A Murder of Crows” from Erlang Solutions) and FLAME to coordinate thousands of users across elastically scaled compute. You’ll learn to write scenarios measuring what actually matters for LiveView performance: mount and event round-trip latencies, diff sizes, and LiveView performance across different LiveView patterns. We’ll explore the architecture, walk through scenario design, and depending on where progress is at, demonstrate spinning up distributed runners against a live application.