Katkat
Back

Design a Direct Messaging System for an E-commerce Marketplace

WhatnotSoftware Engineer · Technical Screen
MediumSoftware EngineerTechnical ScreenSystem Design

Last updated

Design a scalable 1:1 direct messaging platform for an e-commerce marketplace that supports real-time chat, message ordering, read receipts, and offline push notifications across multiple devices.

Scenario

In a live and social e-commerce marketplace, buyers and sellers frequently communicate directly to ask questions about product listings, negotiate prices, coordinate shipping, and resolve disputes. You are tasked with designing the 1:1 Direct Messaging (DM) platform to support this interaction.

Requirements

Functional Requirements

  • 1:1 Messaging: Users can initiate chats and exchange text messages (up to 2,000 characters) in real time.
  • Message Status & Ordering: Messages must display delivery status (sent, delivered, read) and be presented to both parties in strict chronological order.
  • Offline Handling & Push Notifications: If a recipient is offline or disconnected, messages must be stored reliably, and a push notification must be triggered.
  • Multi-Device Sync: A user logged in on multiple devices (e.g., mobile app and web browser) must see an up-to-date conversation state and synchronized read receipts.
  • Listing Context: Messages can optionally be tied to a specific listing ID or order ID for context.

Non-Functional Requirements

  • Low Latency: Real-time delivery latency should be under 200 ms for online users.
  • High Availability & Durability: Zero message loss once acknowledged by the server (at-least-once delivery semantics to the client with deduplication).
  • Scale Assumptions:
    • 20 million Daily Active Users (DAU).
    • Average of 50 million messages sent per day.
    • Peak message generation rate: ~2,000 messages/second.
    • 5 million concurrent WebSocket connections at peak.

What Your Design Should Cover

  1. High-Level Architecture: Protocol choices (WebSockets, HTTP/2, push services), connection management, and routing layers.
  2. Data Model & Storage: Database schema for conversations, messages, and delivery states, including storage engine selection for high-write chat workloads.
  3. Message Lifecycle & Delivery Semantics: Step-by-step flow from sender client to recipient client, handling ACKs, retries, and sequence numbering.
  4. Reconnection & Catch-Up: How clients sync unread messages and missed updates after disconnects or network switches.
  5. Trade-offs: Latency vs. strong ordering consistency, long-lived connection resource usage, and storage partitioning strategies.

Hints

Hint 1

Consider decoupling connection handling (stateful gateway servers) from business logic and message persistence (stateless services) using a message broker or distributed pub/sub.

Hint 2

To guarantee strict message ordering within a conversation, assign monotonically increasing sequence numbers or IDs scoped per conversation rather than relying purely on client or wall-clock timestamps.

Hint 3

For multi-device sync and reconnects, design a pull-based cursor or sequence-based sync endpoint that retrieves all events after the client's last acknowledged sequence ID.

Follow-up questions

  • How would you extend this design to support rich media (images and short video clips) without saturating the real-time WebSocket connection layer?
  • If a popular seller receives thousands of inquiries during a live auction stream, how do you prevent fan-in bottlenecks on that seller's connection and inbox?
  • How would you implement end-to-end spam and fraud detection (e.g., detecting phone numbers or off-platform payment links) in real time without increasing delivery latency significantly?