PAPA: Performance of Android Production Applications
summary
Claims a user‑centric model (MILK) and a toolkit that finds root causes in production: use Android Vitals then add custom production perf analytics and tiny sampled production tracing to get context, reproduce issues, and write tests. Measure real startup and interaction latency with Perfetto traces and macrobenchmark hooks, spot surprises like 1.7 second searches or missed 60fps frames vs a 200 ms interaction threshold, fix by joining queries or changing threading, and verify with repeatable traces.
AI-written summary · about
chapters (11)
- 0:00Start
- 12:04Interaction Latency Thresholds
- 16:15User Centric Performance
- 16:46Universal steps to fix perf issues
- 18:41Information gathering 3 goals
- 19:05Perf feedback loops
- 20:35Compile time / Lint
- 23:44Custom prod perf analytics
- 24:50Prod tracing
- 31:26Keep it real
- 32:01Investigate outside-in
related talks
35mIn Search of Performance: the Subtle Art of Android Profiling
39mDeep-Diving Android Performance with Perfetto
34mClient Tracing At Scale
39mAndroid App Performance in a Nutshell
42mMILK: a user-centric mobile performance model
13mPerfetto Your App for Perfect Performance
40mDebugging Android Performance at Scale: Lessons from Building Datadog's Mobile Profiler
42mPerformance Paradox: Navigating the Complexities of Mobile App Optimisation