Katkat
Back

Design an Instagram-like Feed System

OpenAISoftware Engineer · Technical Screen
MediumSoftware EngineerTechnical ScreenSystem Design

Last updated

This question assesses the ability to architect a high-scale feed system by balancing pre-computation (push model) versus on-demand computation (pull model) to minimize latency for social media users.

Problem Statement

Design an Instagram-like news feed system. Users follow other users, and when they open the application, they see a ranked list of posts from the people they follow.

Functional Requirements

  1. Users can post content (images/text).
  2. Users can follow other users.
  3. Users can view their personalized feed.

Non-Functional Requirements

  1. High read availability: Feed viewing is the most frequent operation.
  2. Low latency: The feed must load in under 200ms.
  3. Scalability: The system must support millions of daily active users and handle "celebrity" users with millions of followers.

Expectations

Provide a high-level design that addresses:

  • Data model: How to store users, follows, and posts.
  • Feed Generation: How to construct the feed (Push vs. Pull approach).
  • Scalability: How to handle the "celebrity" problem (e.g., an account with 10M+ followers).
  • Storage: The trade-offs between SQL and NoSQL databases for this use case.

Example Flow

  • User A follows User B.
  • User B posts an image.
  • User A logs in and fetches their feed. The system must retrieve the most recent posts from all followed accounts, ranked by time and relevance.

Hints

Hint 1

Consider the trade-offs between a 'pull' model (fetching feeds at read-time) versus a 'push' model (pre-computing feeds into a cache upon posting).

Hint 2

How does the performance of a push model degrade when a user with millions of followers posts a new update?

Hint 3

Think about how caching strategies (like Redis) can alleviate the load on your primary persistent storage.

Follow-up questions

  • How would you modify your design to support a ranking algorithm that displays posts based on user engagement rather than simple reverse-chronological order?
  • If we implement a 'push' model, how would you handle the 'celebrity' problem to avoid excessive fan-out writes?
  • How would you ensure feed consistency if a user deletes a post shortly after publishing it?