May 2023 Summaries

Firebase Cloud Messaging’s experience supporting the 2022 World Cup illustrates how major live events can create abrupt, global notification spikes requiring extensive preparation. During the Argentina–France final, FCM delivered more than 115 billion notifications to over 400 million users and saw traffic rise from an 18 million QPS baseline to a 46 million QPS peak within seconds. Effective preparation involves forecasting traffic from historical data, provisioning redundant capacity early, locating infrastructure near traffic sources and destinations, and checking high-percentile dependency latency and availability. The account recommends isolating event traffic from regular workloads, configuring load balancers to distribute sudden bursts across nearby regions, and using layered safeguards such as server-side throttling and per-sender quotas. It also discusses choosing between vertical and horizontal scaling based on workload bottlenecks, and emphasizes realistic load testing on separate infrastructure that reproduces rapid, irregular traffic patterns rather than only sustained peak volumes.

Google I/O 2023 introduced a range of Firebase updates aimed at simplifying full-stack development and integrating generative AI capabilities. New Firebase Extensions based on the PaLM API enable functions such as chatbots, text summarization, language tasks, and secure AI calls, while a new publishing experience allows developers to create and distribute their own extensions. Cloud Functions for Firebase 2nd generation reached general availability with higher memory limits, concurrency, Firestore triggers, and Python support in preview, and Firestore added generic OR queries and expanded IN-query limits. Security updates include App Check protection for Authentication, Unity and C++ game SDKs, and replay protection for sensitive backend actions, while Terraform support enables infrastructure provisioning through code. Firebase Hosting expanded experimental deployment support for Astro, Nuxt, SvelteKit, Flask, Django, and Flutter web, alongside dynamic preview channels for collaborative testing. Additional improvements include faster Android app startup, expanded App Quality Insights in Android Studio, in-app tester feedback through App Distribution, and real-time Remote Config updates for faster feature rollouts and reversions across several platforms.

Crashlytics is automatically introducing a smarter event-grouping algorithm designed to improve how crashes, non-fatals, and ANRs are organized into issues, helping developers identify and prioritize app failures more accurately. The previous approach often relied on a single shared code line, which could split events from the same root cause into duplicate issues or combine unrelated failures into one issue, particularly after line-number changes or when errors passed through common handlers. The updated system evaluates multiple event characteristics, including stack-trace frames, exception messages, error codes, platform details, and error types, while selecting the most relevant frames for analysis. It also adds variants, which are subgroups within an issue that share a failure point and similar stack traces, enabling investigation of potentially distinct root causes. Existing issues will remain unchanged, while incoming events will first be matched against old issues and otherwise processed through the new algorithm, gradually adding new-style issues with clearer titles, expanded metadata, stronger search capabilities, fewer duplicates, and more meaningful alerts.