The Facebook Mobile Performance Tooling Ecosystem: An Overview
summary
Claims that focusing on performance helped reverse declining engagement and that Facebook measures latency and jank with concrete signals: time to recent content (TTRC) and large frame drops defined as four consecutive misses. Tooling includes a device lab, Infer static analyzer, the Profilo tracing library, QPL event logging and Comparison View for trace-level, statistically significant version-over-version diffs. Goals use “percent good” split by device year class to avoid flagship bias. A proposed Performance Insights UI aims to auto-diagnose causes like cache-to-network shifts and guide fixes.
AI-written summary · about
chapters (10)
- 0:00Introduction to mobile performance
- 1:17Why performance matters at Facebook
- 3:20Key performance metrics explained
- 5:25Aggregating and goal-setting strategies
- 6:28Defense vs offense in performance
- 7:51Tooling for development phase
- 9:57Testing with Quick Performance Logger
- 11:20Need for production tooling
- 12:42Debugging with Profilo and Comparison View
- 16:36Future of performance tooling
related talks
13mCreating an inhouse library to measure performance
33mDemystifying Web Performance Tooling
16mFrom Beginner to Expert in Web Performance
50mHow to Think like a Performance Engineer
21mState of the Union for Speed Tooling
16mFrom beginner to expert in web performance
01h 03mDeveloping C++ @ Facebook Scale
33mRedefining Web Performance