System design interview question
Design a Multi-Region Real-Time Notification Delivery System (like Firebase Cloud Messaging at scale)
Design a notification delivery platform that serves 500M monthly active users across 6 continents, handling 2M notifications per second at peak with latency SLAs of <500ms for 99th percentile delivery. The system must support multiple channels (push, email, SMS, in-app), handle device lifecycle events (registration, unregistration, token refresh), provide reliable at-least-once delivery semantics with deduplication, and enable rich targeting (user segments, A/B testing, time-zone aware scheduling). Users expect notifications within seconds of trigger events, and the system must gracefully handle network partitions, device offline scenarios, and traffic spikes during viral campaigns. Address how you'd handle consistency trade-offs between real-time delivery and eventual consistency in distributed state, design for extensibility to support new channels, and ensure cost efficiency given massive scale.
Practice this question step by step with an AI interviewer that reviews your requirements, your architecture and your trade-offs, and scores the design.