Katkat
Back

Design a multi-size storage locker system

AmazonSoftware Engineer · Onsite
MediumSoftware EngineerOnsiteSystem Design

Last updated

This question tests your ability to design a scalable, distributed reservation system that handles resource allocation (bin packing) under concurrency constraints.

Design a Multi-Size Storage Locker System

Design a backend service for a network of automated storage lockers found in public locations (e.g., apartment complexes or transit hubs). The system must allow users to reserve, deposit, and retrieve packages.

Functional Requirements

  • Users can view available locker sizes (Small, Medium, Large) at a specific facility.
  • Users can reserve a locker for a limited duration.
  • The system must track the state of every locker: EMPTY, RESERVED, OCCUPIED.
  • Support package drop-off (transitioning from RESERVED to OCCUPIED) and retrieval (clearing the locker).

Non-Functional Requirements

  • Concurrency: Ensure no two users can reserve the same locker simultaneously.
  • Scalability: The system should support thousands of facilities globally.
  • Availability: Reservations should succeed even if the local kiosk hardware has intermittent connectivity, though the locker state must remain consistent.

Deliverables

  1. API Design: Define the core endpoints for viewing availability, reserving, and confirming drop-off.
  2. Data Model: Design the schema for lockers, facilities, and bookings.
  3. Concurrency Strategy: Explain how you prevent race conditions when multiple users attempt to book the last available locker of a specific size.
  4. System Architecture: Propose a high-level design that handles the distributed nature of the hardware terminals and the central reservation system.

Hints

Hint 1

Consider using database-level locking or optimistic concurrency control with versioning to manage the 'last available locker' problem.

Hint 2

Think about whether a reservation should be a state transition on a locker object or a separate record in a 'Bookings' table.

Hint 3

Address how you would handle a scenario where a user reserves a locker but never shows up, including the state recovery process.

Follow-up questions

  • How would you optimize the system if a facility has thousands of lockers and a high churn rate for reservations?
  • How do you handle 'stuck' lockers where a sensor reports an object is present, but the system shows the locker as empty?
  • If we introduce a multi-tenant model (e.g., different logistics companies using the same lockers), how would your data model change?