Real-Time Chat System

Real-time chat is one of the most frequently asked system design questions at FAANG companies because it combines WebSocket connection management, message ordering guarantees, presence detection, and offline message queuing into a single problem. In this challenge, you will architect a chat system that supports one-on-one messaging, group chats with up to 500 members, read receipts, typing indicators, and online/offline presence. The connection layer uses API Gateway WebSocket APIs with Lambda authorizers for authentication. Messages flow through an SQS FIFO queue to guarantee exactly-once, ordered delivery, then fan out to all group members via SNS. Message storage uses DynamoDB with a composite key design — partition key on conversation ID, sort key on timestamp — enabling efficient pagination of chat history. For presence detection, you will implement a heartbeat pattern using ElastiCache with TTL-based expiration: clients send heartbeats every 30 seconds, and absence of a heartbeat triggers an offline event. The architecture handles the thundering herd problem when a popular user comes online by batching presence notifications. You will also design the push notification path for offline users using SNS and Firebase Cloud Messaging. This challenge teaches connection state management, message ordering in distributed systems, and efficient fan-out strategies.