Android performance monitoring with automation
summary
Process for automated Android performance monitoring that builds a global device cloud with custom boxes (24 devices each), an e-finger for remote maintenance, and packet capture plus 60 fps video and bi‑directional audio to capture user‑side signals. Uses Appium and image tools (AirTest) wired into CI (Jenkins) for scripted user flows, runs ~5 tests/device/hour yielding 40,000 sessions per device per year, and finds most apps can be optimized by ~1s by fixing server/CDN, TLS, caching and client rendering issues.
AI-written summary · about
chapters (11)
- 0:00Introduction to performance monitoring
- 3:19High fidelity user data approach
- 6:07Building a global device cloud
- 10:03Mapping the user experience genome
- 17:24Hidden signals and root causes
- 18:49Analyzing user session examples
- 21:34Automating user flow testing
- 28:19Open source automation tools
- 33:13Managing flaky test automation
- 36:46Scaling automation with Headspin
- 37:31Q&A summary
related talks
41mDevice Server for an Android Farm or Running More Than 30k Tests Every Day
39mDeep-Diving Android Performance with Perfetto
29mPros and Cons of Device Farms Available for Android
39mAndroid App Performance in a Nutshell
43mAndroid Apps Testing in 2019
49mAutomated Solutions For Non-Functional Requirements
26mAndroid CI: 2016 edition
33mAndroid Benchmarking and other stories