Katkat
Back

Design a Scalable News Feed

AmazonSoftware Engineer · Technical Screen
EasySoftware EngineerTechnical ScreenSystem Design

Last updated

This problem evaluates your ability to design a feed generation architecture, specifically comparing push vs. pull models for users with varying follower counts.

System Design: Scalable News Feed

Design a backend service that generates a personalized 'Home Feed' for a social media application. Users can create posts, follow other users, and view a feed consisting of the latest posts from those they follow.

Requirements

Functional Requirements:

  1. Users can post content (text/images).
  2. Users can follow/unfollow other users.
  3. Users can view their feed, which displays posts from followed users in reverse chronological order.

Non-Functional Requirements:

  1. High availability and low latency for feed retrieval.
  2. The system must support at least 100 million daily active users.
  3. Feed updates should appear within a few seconds of a post being published.

Scope

Your response should cover:

  • High-level architecture and API design.
  • Data schema for users, follows, and posts.
  • The mechanism for feed generation (Push vs. Pull strategy).
  • Handling of 'Celebrity' accounts (users with millions of followers) to prevent system bottlenecks.
  • Caching strategy to optimize read performance.

Hints

Hint 1

Consider the trade-offs between 'Fan-out on write' (push) and 'Fan-out on load' (pull) strategies based on a user's follower count.

Hint 2

Think about how to prevent a celebrity account's post from triggering millions of simultaneous database writes when they publish a new update.

Hint 3

Decide whether to store the feed in a relational database or a NoSQL key-value store, and why one might be better for high-read throughput.

Follow-up questions

  • How would you modify the system to support a 'ranked' feed based on engagement rather than strictly reverse-chronological order?
  • If a user deletes a post, how do you ensure it is removed from all feeds in a distributed system?
  • How would you handle the storage and delivery of media content associated with posts to minimize latency?