Can Elixir Bring Down Phoenix?
summary
Title says it all: Elixir and Phoenix power a high‑scale media backend and the talk gives concrete claims, tools and numbers. Production scale is 15M app installs, 1.5B pageviews and 3B push notifications a month with a peak of 225M pushes in three hours. Techniques include ghosting traffic to stage, a homegrown Locust library for chaotic load, ETS caching, Task.start and Hackney tuning. Result: baseline app sustains 8x traffic without extra servers until the DB becomes the bottleneck; ETS cuts response time roughly in half.
AI-written summary · about
chapters (8)
- 0:00Introduction and background
- 1:36Load testing and performance
- 4:59Duplexing traffic for testing
- 7:27Multiplying traffic for stress
- 9:55Locust library and chaos
- 13:55Optimizing with ETS
- 17:22Implementing ETS caching
- 22:55Demo and conclusion
related talks
38mElixir Adoption at Scale: A Comprehensive Guide
45mElixir: une potion pour mixer complexité et concurrence en temps réel
44mElixir & Phoenix - Fast, Concurrent and Explicit
01h 00mPhoenix a Web Framework for the New Web
19mElixir – what it is and why is it amazing?
32mWhy I choose Phoenix
41mElixir & Phoenix – fast, concurrent and explicit
45mPhoenix a web framework for the new web