Katkat
Back

Design several large-scale systems

RipplingSoftware Engineer · HR Screen
HardSoftware EngineerHR ScreenSystem Design

Last updated

This question tests the ability to design a distributed, high-concurrency reservation system that requires strong consistency for inventory management and low-latency read access for search.

System Design: Global Event Booking Platform

Design a global event booking system (e.g., for concert tickets or sports matches) that handles high-traffic spikes when tickets first go on sale. Millions of users will attempt to purchase a limited number of tickets simultaneously.

Requirements

Functional:

  1. Users can search for events by location, date, and category.
  2. Users can view seat availability in real-time.
  3. Users can reserve seats for a temporary period (e.g., 10 minutes) before completing payment.
  4. Ensure that a single seat cannot be sold to two different users.

Non-Functional:

  1. High availability and low latency for seat selection.
  2. Strong consistency for booking operations to prevent overselling.
  3. Scalability to handle sudden traffic surges (thundering herd problem).

Scope

Provide a high-level architecture including:

  • API design for seat selection and booking.
  • Storage schema choice (SQL vs. NoSQL) and why.
  • Strategies for handling concurrency during the "check-out" flow.
  • Handling the read-heavy load of users checking seat maps simultaneously.

Hints

Hint 1

Consider the trade-offs between using a relational database with row-level locking versus an in-memory distributed store for the reservation state.

Hint 2

Think about how to offload read-heavy traffic from the primary booking database using caching or read replicas.

Hint 3

Address the 'thundering herd' issue by implementing a virtual waiting room or using rate-limiting strategies for initial traffic spikes.

Follow-up questions

  • How would you modify your design if you needed to support 'dynamic pricing' that changes based on real-time demand?
  • What changes would be necessary to ensure the system remains available for search even if the booking database is undergoing maintenance or is partially unavailable?
  • How would you handle a distributed transaction failure where a user pays for a ticket, but the inventory update fails due to a network partition?