Course contentsShow
← System Design
Lesson 1648 of 1,91940. Real-World System: Social Media FeedFree lesson

Hybrid Fanout Strategy

Combining push and pull models by using fanout-on-write for most users and pull for celebrities with millions of followers.

Hybrid Fanout Strategy

What you'll learn: How to combine push and pull fanout models to handle both regular users and celebrities efficiently in a social media feed system.


The Core Idea

A hybrid fanout strategy intelligently switches between fanout-on-write (push) and fanout-on-read (pull) based on user characteristics, typically follower count. The key insight: most users benefit from pre-computed feeds (push), but celebrities with millions of followers would overwhelm the system if we tried pushing their posts to everyone at write-time.

How It Works

When a user publishes a post, the system checks their follower count:

  • Regular users (< threshold): Use fanout-on-write. The post is immediately pushed to all followers' pre-computed feeds. Fast reads, manageable writes.

  • Celebrities (> threshold): Use fanout-on-read. The post is stored centrally, and followers' feeds pull from celebrity accounts at read-time. This avoids writing millions of records.

  • Your feed: A blend of pre-computed posts (from regular users you follow) plus real-time queries (for celebrities you follow).

Real-World Example

Imagine you follow 500 people: 490 regular users and 10 celebrities. Your feed contains:

  • 490 pre-pushed posts sitting in your timeline cache
  • 10 celebrity posts fetched on-demand when you open the app

The system merges and sorts these sources, presenting a unified feed.

The Threshold Decision

The "celebrity threshold" (e.g., 1 million followers) is tunable based on:

  • Write capacity
  • Acceptable fanout latency
  • Storage costs

Key Takeaway: A hybrid fanout strategy combines the read-speed benefits of push for regular users with the write-efficiency of pull for high-fanout accounts, solving the celebrity problem while maintaining feed performance for most users.