System Design

Case Study: WhatsApp

Requirements

Functional

• 1-to-1 messaging
• Group chat (256 members)
• Message delivery status
• Online/offline status
• Media sharing

Scale

• 2B users
• 100B messages/day
• Real-time delivery

Core: WebSocket Connections

Client ←→ WebSocket Server
          ↓
      Message Queue
          ↓
      Chat Server
          ↓
      Database

Message Flow

1. User A sends message
2. WebSocket Server → Message Queue
3. Chat Server:
   - Save to database
   - Check if User B online
   - If online: Push via WebSocket
   - If offline: Queue for later (Push notification)
4. Delivery confirmation

Database Schema

-- Messages
CREATE TABLE messages (
  id BIGSERIAL PRIMARY KEY,
  from_user_id BIGINT,
  to_user_id BIGINT,
  content TEXT ENCRYPTED,
  created_at TIMESTAMP,
  delivered_at TIMESTAMP,
  read_at TIMESTAMP
);

-- Sharding key: user_id % 4096

Key Design Decisions

1. End-to-End Encryption:

• Messages encrypted on device
• Server can’t read content

2. Connection management:

• Each user → 1 WebSocket connection
• Server tracks: user_id → connection_id

3. Offline messages:

• Store in database
• Push notification
• Deliver when user comes online

4. Group chat:

• Message replicated to all members
• Last seen per member

Scaling

2B users, 50M concurrent connections:

• 1 server = 10K connections
• Need 5,000 servers
• Load balancer with sticky sessions

Keyingi dars: Uber loyihalash.