Google Flights

January 15, 2023 1:30 PM PST

---

This document summarizes a mock system design interview focused on Google Flights, targeting a senior engineer level. The interview covered functional and non-functional requirements, high-level design, API specifications, and database schema considerations.

Requirements

Functional Requirements

• Customers can search for flights by source, destination, and date on the Google website.
• Customers can select a flight and proceed to booking (reservation and payment processes are out of scope).
• Customers can filter and sort flights based on price, duration, and number of stops.
• Customers can track historical prices for a given flight.

Non-Functional Requirements

• Availability
• Eventual consistency
• Scalability

High-Level Design

• User Interface: Entry point for users to access flight services.
• Flight Management: Queries the flight database.
• Data Ownership: Google owns the service but relies on third-party flight data services for information retrieval.

Data Retrieval

• Data Model: A pull model can be used to retrieve data from third-party services.
• API Design:
  • Query API:
    • Request:
      • Source: string
      • Destination: string
      • Start date: timestamp
      • Stop: boolean
    • Response: A list of flight instances.

Flight Instance Structure

• Flight Instance ID
• Start Date
• End Date
• Flight ID
• Flight Company
• Stops
• Duration

Error Handling

• Differentiate between internal errors and no available flights.
• Exceptions:
  1. 500 Internal Errors
  2. 400 Retriable Errors

Handling Stops

• Filter conditions for 1, 2, or 3 stops can be implemented.

Database Schema

• Flight Table:
  • Flight ID
  • Flight Company
  • Start Time
  • End Time
  • Departure
  • Arrival
  • Duration
• Flight Instance Table:
  • Departure Time
  • Status
  • List of Tickets
    • Ticket:
      • Fare
      • Type
      • Remaining Number

Route Calculation

• Routes can be computed in real-time or pre-computed for popular flights.
• Criteria for computation based on search patterns.

Integration with Third Parties

• Push Model: Flight companies can push updates to Google when changes occur.
• Pull Model: Google can periodically pull data from third-party APIs.

Availability and Consistency

• Strong consistency is not required; NoSQL databases like DynamoDB can be used for scalability.
• A caching layer can be implemented for high availability.

Interview Feedback

• The candidate demonstrated potential for a senior engineer role but did not meet the criteria for L5.
• Areas for improvement include:
  • Requirement gathering, especially regarding scale.
  • Communication during diagramming and deeper exploration of design details.
  • API and database schema design should be more detailed and timely.

Audience Feedback

• The interviewee struggled with understanding the interviewer's concerns and questions.
• Emphasis on the importance of confirming basic requirements and the need for a structured approach to design discussions.

Technical Considerations

• The system must handle a significant volume of data, including 40,000 airports and 100,000 flights.
• Considerations for using SQL vs. NoSQL databases and the implications for data retrieval and storage.
• Strategies for handling sold-out flights and ensuring data consistency across systems.