[Chat App] Architecture - High Level Design

Here are the infrastructure components and their relationships and responsibilities:

Infrastructure components

⚙️ Compute & Routing

• HTTP Load Balancer：Distributes REST API traffic evenly across HTTP servers to ensure scalability and high availability.
• WebSocket Load Balancer：Balances long-lived WebSocket connections, helping the system handle many concurrent users reliably.
• EC2 Instances (HTTP)：Handle stateless API operations like user creation and message history retrieval, allowing horizontal scaling.
• EC2 Instances (WebSocket)：Maintain persistent connections for real-time features like sending/receiving messages, ensuring low-latency interaction.

🫙 Data Storage

• DynamoDB：Provides fast read-write, scalable storage for various chat-related data: users, rooms, messages, and user-room relationships—all with flexible schema support and high throughput.

⚡️ Real-time Communication

• Redis：Stores WebSocket connection IDs and enables cross-server message broadcasting between WebSocket EC2 instances.

Component responsibilities

1. HTTP GET /users endpoint：
   Client sends a request to retrieve all user data. The backend queries User DynamoDB and returns the result.

2. HTTP POST /users endpoint：
   Client creates a new user. The backend processes the data and writes it to User DynamoDB.

3. HTTP GET /rooms/{room_id}/messages endpoint：
   Client requests the message history of a chat room. The backend fetches the messages from messages DynamoDB and returns them.

4. WebSocket create_room action：
   Client sends a command to create a chat room via WebSocket. The backend stores the user-to-room relation in user_rooms DynamoDB.

5. WebSocket create_room (continued)：
   The system also creates the room metadata and stores it in rooms DynamoDB.

6. WebSocket create_message action：
   Client sends a new message via WebSocket. The backend stores it in messages DynamoDB.

7. Redis Broadcast：
   The system uses Redis to store WebSocket connection IDs and serve as a broadcast intermediary. This allows different WS EC2 nodes to be aware of existing connections, enabling real-time message forwarding to users, even if they are connected to different servers.