From 0 to 100: Fine-tune all the things!
summary
Mapbox Navigation SDK performance work builds an automated pipeline that runs UI tests on AWS Device Farm, captures Systrace and adb dumpsys cpuinfo output via custom test @Rules, parses noisy dumpsys text with Python, and stores results in a data warehouse. Control routes take 2–10 minutes (avg ~5), 30+ runs cost ~2–2.5 hours, Java heap moves from ~6.5 to ~7 MB, and production telemetry adds real‑world coverage at the cost of careful low‑overhead instrumentation.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:09Poor performance
- 2:43Maps
- 4:12Routing
- 6:14Route Processing
- 8:46Performance Areas
- 18:18Custom Test Rolls
- 20:00Reproducible Test
- 21:25Data Management
- 23:51Device Farm
- 25:04Drawbacks
- 25:57Time
- 26:31Documentation
- 28:00Unreliable data
- 28:39Production telemetry
- 31:32Process walkthrough
- 35:16Whats next
related talks
30mLearn from developers: Lyft and Google Maps share their experiences optimizing app performance
21mStreamlining Android app development and release processes
37mPerformance Monitoring- 0 to 60 in 40 Minutes
42mDebugging Application Performance
40mAndroid performance monitoring with automation
27mIntroducing the Mobile PerfSDK
32mGet a performance score for your app
47mBuilding with the Navigation UX Framework on Android