How OLX’s New Chat Became 75% Cheaper and Saw 40% Traffic Growth
summary
The talk explains how OLX rebuilt its chat using Elixir and Ejabberd to cut costs by 75% and support 40% more traffic by improving resource utilization and stability. It describes an incremental, modular, and extensible migration process, covering technology choices and measures to avoid user impact during the transition.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 1:13Old chat problems
- 3:20New tech choice
- 4:45Why Erlang and Elixir
- 8:16Actor model and concurrency
- 10:20Supervision and fault tolerance
- 14:14Fault-tolerant system
- 17:45System evolution difficulty
- 18:50Hook-based architecture
- 21:19Mercúrio born
- 22:03Scaling with BEAM
- 23:48Monolith to microservices
- 26:13Data migration
- 28:57Monitoring and metrics
- 32:06Unified frontend
- 34:13Platform integration
- 35:17Results achieved
- 36:18Pending challenges
- 38:27Database and scalability
- 40:59DB questions
- 43:06Productivity questions
- 45:58Language adoption questions
- 47:47Closing
related talks
01h 52mPurple Carpet
49mKeynote: How to sell Elixir (Again)
38mElixir Adoption at Scale: A Comprehensive Guide
32mDigitalizing the Steel Industry with Elixir
49mAn Introduction to Erlang From behind the Trenches
45mElixir: une potion pour mixer complexité et concurrence en temps réel
43mThe River
44mIdioms for Building Fault-tolerant Applications with Elixir