App Performance Optimization

1 post · Dec 2019 – Mar 2026

What this trail covers

Grab has focused on improving app performance, specifically Time To Interactive (TTI), by optimizing app startup time and service tile rendering. Efforts included implementing tile caching, removing non-essential app start animations, adopting static linking on iOS to reduce binary loading time, and utilizing coroutines for parallel library initialization on Android. The team also enhanced their visualization and tracking of performance metrics using Metabase and migrated to an in-house experimentation platform.

All Grab trails

Mar 2026Enabling R8 optimization at scale with AI-assisted debuggingThis post details the successful enablement of R8 optimization for the Grab Android app, addressing systemic performance challenges and ANR rates. It introduces AI-assisted debugging using custom MCP tools and GitLab CI integration to accelerate the investigation and resolution of R8-specific issues in obfuscated code. The solution involves automatic APK decompilation, stack trace deobfuscation, and AI-driven parallel merge request creation for testing multiple solutions simultaneously. This approach significantly reduces the manual effort and time required for debugging, enabling faster iteration and deployment of R8 optimizations.
Jan 2021Serving Driver-partners Data at Scale Using Mirror CacheIntroduced a 'mirror cache' solution for the Drivers Data service to improve the serving of driver-partner data. This involved replacing the previous cache-aside strategy with a distributed in-memory cache (Ristretto) that replicates data across all nodes in the cluster. The mirror cache aims to increase the local cache hit ratio by ensuring that data fetched by one node is available to others, reducing redundant calls to Redis and MySQL. Key components include an in-memory cache, a replicator with a membership store, notifier, and gRPC server, and a design that supports auto-scaling and eventual consistency.
Dec 2019Journey to a Faster Everyday Superapp Where Every Millisecond CountsThis post details the optimization of the Grab passenger app's Time To Interactive (TTI) by focusing on app startup time and service tile rendering. Key improvements include implementing tile caching (saving ~3s), removing app start animations (saving ~1s), adopting static linking on iOS (reducing pre-main time by 41% and overall dynamic frameworks by 88%, saving ~0.8s with CocoaPods merge), and utilizing coroutines for parallel library initialization on Android (saving ~0.4s). The team also improved visualization and tracking of performance metrics using Metabase and migrated to an in-house experimentation platform, leading to significant improvements (~2.5s on Android, ~0.5-0.7s on iOS) by removing a third-party analytics library.
Advertisement