Designing a Chat Application

A chat system enables real-time messaging between users at scale.

Requirements

• 1-on-1 and group messaging
• Real-time delivery
• Message history
• Online/offline presence
• Read receipts

Architecture

Clients ↔ WebSocket Servers ↔ Message Service
                                    ↓
                               Kafka
                                    ↓
                           Message Store (Cassandra)

Key Design Decisions

Connection Management

Use WebSockets for persistent bidirectional connections. Each server handles ~100K concurrent connections.

Message Storage

Cassandra is ideal — optimized for time-series, high write throughput.

Schema: (chat_id, message_id, sender, content, timestamp)

Message Fan-out

For group chats:

1. Message arrives at server
2. Kafka publishes to topic
3. All group members' servers consume and push

Presence System

Redis stores online status with TTL:

SET user:123:online 1 EX 30

Heartbeat from client resets TTL.

Delivery Guarantees

• Message ID generated at server (UUID)
• Client ACK confirms delivery
• Retry for unacknowledged messages

Conclusion

Chat systems require WebSockets for real-time, Cassandra for message storage, and Kafka for fan-out at scale.