Why should you care about coldstart and how?
summary
Concrete definition: cold start ends when the user sees useful data, so Yahoo Finance sets a 2,000 milliseconds median goal and measures phases with a phase stopwatch and cold start register. Tooling: local telemetry pipes phase timestamps to Splunk and a Google-timeline visualizer; CI runs a Gradle task with a local HTTP server with NanoHTTPD plus Stetho to mock APIs and catch regressions. Optimization example: moving ticker fetch earlier saves ~300 ms and production cold-start incidents drop by a large factor.
AI-written summary · about
chapters (11)
- 0:00Introduction to cold start
- 1:37User expectations for app speed
- 4:44Defining useful app startup
- 8:33Setting performance goals
- 12:06Measuring cold start phases
- 16:58Visualizing performance data
- 23:42Building a telemetry system
- 26:10Alerting for performance issues
- 28:59CI regression testing
- 35:06Key optimization lessons
- 38:13Q&A on app performance
related talks
24mHow to Improve App Startup Time
36mWin on startup time! Improving app launch performance
24mImproving the Bootup Speed of AOSP
45mDeep dive into Modern Android Performance
18mHow we improved our app startup time by 50%
37mPerformance Monitoring- 0 to 60 in 40 Minutes
05mWorkshop: Inspect and improve app performance
44mHow to make it stop lagging, please