Should I use fan-out on write or fan-out on read?

Hybrid. Fan-out on write for users with under 10K to 100K followers. Skip fan-out for high-follower accounts and pull their tweets at read time. This gives you fast reads without the celebrity blowup.

How do you handle a tweet from someone with 100M followers?

Skip fan-out on write. Store the celebrity's recent tweets in a hot read-through cache. Timeline reads merge the per-user precomputed cache with the celebrity cache. The hot key is isolated and scales reads independently.

How does Twitter search work in near real time?

A Kafka stream feeds an Elasticsearch indexer (or Twitter's Earlybird, a custom Lucene fork). Tweets are indexed within seconds. Search queries hit a memory tier first, then on-disk segments. Ranking uses a learned model over recency and engagement.

What is the data model for the follow graph?

Two denormalized tables: followers (lookup followers of a user) and following (lookup who a user follows). Both sharded by the primary user_id. Sometimes stored in a graph DB for richer queries.

How are trending topics computed?

A Flink job aggregates hashtag counts per region per minute, then computes velocity (current rate divided by 24-hour rolling baseline). Topics with the highest velocity surface as trending.

System design interview guide

Design Twitter: System Design Interview Guide

Short answer

Twitter system design centres on the home timeline. For most users, a new tweet is copied into each follower's timeline cache when it is posted (fan-out on write). For celebrities with millions of followers this is skipped, and their tweets are fetched and merged when the timeline is read. Search and trending topics run on the same tweet stream.

in this guide

Design Twitter, part by part

7 parts open to everyone · 3 with a free account · 3 with the course

You can read 7 of 13 parts. A free account opens the deep dive; the course opens all of it.

Requirementsopen
Scale estimatesopen

Why this question is asked

Design Twitter forces you to confront one of the cleanest trade-offs in system design: fan-out on write vs fan-out on read. It also tests how you handle hot keys (celebrities with millions of followers), real-time ranking, and search at scale.

This page is the free part.

The course goes deeper on the Twitter design

₹1,111 in India$25 everywhere elseonce, for the whole course

The System Design course covers the Twitter design across a run of lessons, not one page. These 4 alone are about 95 minutes of step-by-step reading, every one with a quiz.

Requirements

Always clarify these in the first 5 minutes of the interview. Do not start drawing boxes until both lists are agreed.

Functional requirements

Users post tweets up to 280 characters, with optional media
Users follow other users
Users see a home timeline of tweets from followed accounts
Users can like, retweet, reply, and quote tweet

Go deeper

This guide is built on these 6 lessons. Each one teaches one piece of the Twitter design above, step by step.

Practice with 770 system design lessons

Lifetime access for ₹1,111 in India or $25 elsewhere. Interactive diagrams, quizzes, and 20 capstone projects to practise on.

Get lifetime access See all interview problems

Two courses by the author of this page:

System Design

770 lessons · 18 free to read

₹1,111 in India · $25 elsewhere, once

Get System Design

You own this course

Generative AI and Agentic AI

214 lessons · 10 free to read

₹1,333 in India · $35 elsewhere, once

Get Generative AI and Agentic AI

You own this course

Every lesson in both courses has an AI tutor beside it. It reads the same lesson you are reading and answers your questions from it. Try the tutor

Save your progress and get the interview checklist PDF

System Design Interview Cheat Sheet: the 7 interview stages, the 10 points a scorecard grades, 12 classic systems. 18 pages. Free with an account.

Twitter, now X, serves 500 million tweets per day to 600+ million users, with home timelines that must update in under 5 seconds.

Designing Twitter is the canonical timeline generation problem. You decide between push-based fan-out (precompute every follower's timeline) and pull-based aggregation (build on read), and the answer is almost always a hybrid that depends on follower count. It also touches search, ranking, and trending topics.

Roni Das

11 June 2026 · 8 min read

Where it shows up

Commonly asked at Meta, Google, Amazon, X (formerly Twitter), LinkedIn, Pinterest, Reddit, and TikTok. It is one of the top 3 system design interview questions for any social product.

Spec sheetTwitter

01Daily active users: 250M
02Tweets per day: 500M
03Tweets per second (peak): 50K
04Average followers per user: 200

worked through below, with the maths

Data modelfree account

Deep dives (4)free account

Unlock with the course · $25One payment, lifetime access to every lesson.

The hard parts of this design are taught step by step in the System Design course: Fan-Out/Fan-In, Database Sharding and Consistent Hashing.

Fan-Out/Fan-In

The fan-out/fan-in pattern splits work across parallel workers and combines the results. Includes a search across shards example and a sequence view.

20 min · diagrams · quiz

Database Sharding

Splitting a database across multiple servers, the nuclear option for scaling writes, and the architecture behind Instagram, Uber, and YouTube.

35 min · diagrams · quiz

Consistent Hashing

The algorithm that makes adding or removing nodes cheap, an interview favorite and production essential.

25 min · diagrams · quiz

Key-Value Stores

How key-value databases work, how to design good keys, and how Redis, DynamoDB and etcd compare. Includes a diagram of a GET and SET operation.

15 min · diagrams · quiz

all part of

System Design Masterclass

770 lessons · about 287 hours · 18 free to read

₹1,111in India, by UPI

$25everywhere else, by PayPal

Get the System Design course Read 18 lessons free first

You own this course

Continue with Fan-Out/Fan-In

pay once,
yours for life

Live from the course bench

Before you design Twitter, try the questions it rests on.

Real interview questions, live diagrams and measured lab results, pulled at random from the lessons. A new set every time you come back.

Two engineers sketching a system design on a whiteboard

where the work happens

Stuck on a card?

Ask the course tutor. It answers with a real diagram, drawn for your question.

Ask the tutor

Laying out the cards…

Users can search tweets by keyword, hashtag, and user

Trending topics surface in real time per region

Notifications fire for mentions, likes, retweets, follows

Non-functional requirements

Timeline read latency under 200 ms at the 99th percentile
Tweet visible to followers within 5 seconds
99.99% availability for read paths
Eventual consistency is acceptable for timelines and like counts
Scale to 500M tweets per day
Search must surface tweets within seconds of posting

Back-of-envelope scale estimates

Show your math. Pulling numbers from thin air signals you have not thought about the load.

Daily active users

250M

Roughly 40% of 600M registered users active daily based on public reporting.

Tweets per day

500M

Average 2 tweets per DAU. Peak rate during major events (sports finals, elections) is ~10x.

Tweets per second (peak)

50K

500M per day with a 10x peak factor.

Average followers per user

200

Median is much lower (~30), but power users with millions of followers skew the average. Used for fan-out cost.

Timeline reads per second

300K

Reads dominate writes by ~6x. Every app open and every pull-to-refresh is a read.

High-level architecture

When a user posts a tweet, the Tweet Service writes to a sharded SQL store partitioned by user_id and emits an event to Kafka. A Fan-Out Service consumes the event. For normal users (few followers), it precomputes by writing the tweet_id into each follower's timeline cache (Redis sorted set, key user_id, score timestamp). For celebrities (millions of followers), it skips fan-out entirely. When a user requests their home timeline, the Timeline Service reads the precomputed Redis cache for normal-follow content and separately fetches the latest tweets from any celebrities they follow, then merges and ranks before returning. Search runs through a parallel Search Service backed by Elasticsearch, which consumes the same Kafka stream and indexes tweets. Trending topics are computed by a Flink job over the tweet stream.

In a real interview, sketch this on the whiteboard before diving into any single box.

Core components

Walk through each service. The interviewer wants to hear what each one owns, not just the names.

Tweet Service

Owns tweet writes. Persists to sharded SQL (Manhattan in the original Twitter stack, or PostgreSQL shards). Emits a TweetCreated event to Kafka. Reads by tweet_id come from a cache fronting the SQL store.

Follow Graph Service

Stores the follower and following relationships. Heavy-read service. Backed by FlockDB historically, or a custom graph store on top of sharded MySQL. Aggressively cached.

Fan-Out Service

Consumes TweetCreated events. For each follower (up to a threshold like 100K), writes the tweet_id into the follower's timeline cache. For high-follower accounts (celebrities), skips fan-out and relies on the Timeline Service to pull at read time.

Timeline Cache

Per-user Redis sorted set keyed by user_id with tweet_id sorted by timestamp. Capped at the most recent 800 tweets per user. Hit rate is over 99% for active users.

Timeline Service

Reads the user's timeline cache, layers in pulled tweets from celebrity follows, ranks (chronological by default, or by relevance), and returns to the client.

Search Service

Elasticsearch or a custom search engine (Twitter's Earlybird is a custom Lucene fork). Indexes tweets within seconds of creation. Serves keyword, hashtag, and user search.

Trending Service

Flink or Storm job over the tweet stream, grouping by hashtag and region. Surfaces top-N trending per locale every minute.

the deep dive

Data model

tweets. Sharded by author_id. Snowflake IDs encode timestamp so they sort naturally. Read by tweet_id is heavily cached.

free with an account

Every table for Twitter, its keys, and which store fits it and why.

One tap with Google. Free, no card.

the deep dive

Deep dives

Push vs pull for timeline generation. Push (fan-out on write) writes a tweet into every follower's timeline cache when the tweet is created. Reads are O(1): just read the cache. But for a celebrity with 100M…

free with an account

4 deep dives on the parts that break at scale: Push vs pull for timeline generation; Handling the celebrity problem (hot keys); Real-time search with Elasticsearch or Earlybird.

One tap with Google. Free, no card.

the deep dive

Trade-offs to discuss

Push vs pull vs hybrid for timelines. Pure push fails for celebrities. Pure pull is too slow for normal users. Hybrid (push for low-follower, pull for high-follower) gets you 99% of the speed without the…

free with an account

5 decisions an interviewer expects you to defend, each with the alternative and why it loses here.

One tap with Google. Free, no card.

for course owners

What the interviewer asks next

A Redis node holding a few million timeline caches dies. What happens to those users, and how do you recover?

with the System Design course

4 follow-up questions with model answers, the interview scorecard for this question, and a mock interview on Design Twitter with the AI tutor.

Unlock with the course · $25One payment, lifetime access to every lesson.

How Twitter actually does it

Twitter (now X) historically used Manhattan, a custom multi-tenant KV store, for tweet and user data. The fan-out service was originally Rockdove, a Scala service that wrote into Redis. The search engine is Earlybird, a custom Lucene fork. The follow graph was stored in FlockDB. The recommendation timeline (For You) uses a system called Heron for stream processing and a deep learning ranking model. Many of these systems have been rewritten or replaced under X.

Sources

Was this helpful?

the question most readers have next

The interviewer's next question: what breaks when the Twitter design runs at full scale?

A whiteboard answer gets you through the first round. The follow-ups are about the parts that fail under real load. The lessons below teach them one at a time, starting with "Fan-Out/Fan-In".

Where this is used in industry

The real examples, each with its source, are on this page: How Twitter actually does it.

The lessons that teach the next part

The fan-out/fan-in pattern splits work across parallel workers and combines the results. Includes a search across shards example and a sequence view.
Fan-Out/Fan-InSystem Design course20 min
Splitting a database across multiple servers, the nuclear option for scaling writes, and the architecture behind Instagram, Uber, and YouTube.
Database ShardingSystem Design course35 min
The algorithm that makes adding or removing nodes cheap, an interview favorite and production essential.
Consistent HashingSystem Design course25 min
How key-value databases work, how to design good keys, and how Redis, DynamoDB and etcd compare. Includes a diagram of a GET and SET operation.
Key-Value StoresSystem Design course15 min

Stuck on a step? Every lesson has an AI tutor beside it. Ask it your question and it answers from the lesson you are reading. Try the tutor

Get the free System Design Interview cheat sheet (PDF)

Where this is taught

System Design Masterclass

770 lessons · about 287 hours · 18 free to read

₹1,111 in India by UPI$25 elsewhere by PayPal, once, for life

See the System Design course

You own this course

Frequently asked questions

India: UPI, cards or netbanking through Razorpay. Elsewhere: card or PayPal. Download at once from My packs.

NEWSystem Design Hands-On

Included in the System Design course: real systems, built live

Postgres Sharding: Complete Hands-On (PgBouncer + Citus). 2 h 36 min and 20 chapters: one PostgreSQL loaded until it breaks, then four shards behind a router you write, PgBouncer and Citus, with the 139-page slide deck. The first 15 minutes are free to watch.

Watch the free 15 minutes See System Design Hands-On

Free PDF · 79 pages

25 real tech interview questions, answered in full

Five from each of our AI, GenAI, Inference, Maths for ML and Forward Deployed packs. Each with a diagram and a model answer. Free by email.

Also by the same author

214 lessons · about 224 hours · 10 free to read

₹1,333 in India by UPI$35 elsewhere by PayPal, once, for life

See the Generative AI and Agentic AI course

what a reader said

It is really nice and good Please join the course

Gyan Kumar, Analyst, Bangalore · System Design Masterclass · read 3 lessons · all reviews

Design Twitter: System Design Interview Guide

Design Twitter, part by part

Why this question is asked

The course goes deeper on the Twitter design