Katkat
Back

Design bookstore and chat messaging systems

DatabricksSoftware Engineer · Onsite
MediumSoftware EngineerOnsiteSystem Design

Last updated

This question tests the ability to design two distinct systems: a reliable transactional e-commerce inventory service and a low-latency real-time chat application, focusing on appropriate database choices and communication protocols.

System Design: Bookstore & Real-time Chat

You are asked to design two microservices for a large online bookstore platform: the Book Inventory service and the Customer Chat service.

Part 1: Book Inventory Service

Design a system to track millions of book titles, their authors, and current stock levels across multiple warehouses.

  • Requirements: High consistency for stock decrements during checkout. Support for searching books by title or author.
  • Scale: 10 million SKUs, 5,000 requests per second.

Part 2: Customer Chat Service

Design a service that allows customers to chat with support agents in real-time.

  • Requirements: One-on-one messaging, delivery status indicators (sent/delivered/read), and message history retrieval.
  • Scale: 100,000 concurrent active users.

Expectations

For both systems, provide:

  1. API Design: Define key endpoints (e.g., checkout_book, send_message).
  2. Data Model: Choose appropriate database technologies (SQL vs. NoSQL) and schema designs.
  3. Architecture: Describe the high-level flow of requests, including load balancing, caching, and state management.
  4. Trade-offs: Explain how you handle system failures (e.g., what happens if a chat message is sent while a user is offline?).

Hints

Hint 1

For the inventory service, consider how ACID properties are essential for preventing overselling during checkout.

Hint 2

For the chat system, analyze why WebSockets or long-polling are preferred over standard HTTP request-response patterns for real-time delivery.

Hint 3

Consider utilizing different storage engines for static catalog data (like book titles) versus high-velocity transient data (like chat messages).

Follow-up questions

  • How would you implement a 'global search' feature that allows users to find books across all warehouses with low latency?
  • If the chat service experiences a sudden spike in traffic, how would you ensure that the message history retrieval does not crash the database?
  • How would you handle eventual consistency if you moved parts of the inventory system to a multi-region distributed setup?