Instagram Architecture

A photo and video feed that stays fast as the user base grows by orders of magnitude.

On the roadmap:Software Architect · step 5: Learn from the giants

Overview

Instagram is a good example of scaling a monolith before splitting it: keep the application stateless, shard the database, cache aggressively, and push media to a CDN. This summary is written from public engineering talks and blog posts; see the linked resources for the originals.

Architecture highlights

Started as a small Django and PostgreSQL service and scaled by adding caching and sharding
Media is stored in object storage and served through a CDN
Feeds mix precomputed timelines with ranking at read time
Heavy use of Memcached-style caching in front of the database

Architecture diagram

Key components

Application servers
Stateless web and API tier that scales horizontally behind load balancers
Sharded relational database
Stores users, relationships and media metadata, split across many shards
Cache tier
Absorbs most reads for profiles, counts and feed fragments
Media storage and CDN
Holds originals and resized variants and delivers them near the viewer

Challenges and trade-offs

Generating unique ids across shards
Ids must be unique and roughly time-ordered without a single bottleneck.
Feed fan-out
Accounts with millions of followers cannot write to every follower's timeline, so they are handled at read time.
Hot keys and cache stampedes
Viral posts concentrate traffic on a few cache keys, which needs request coalescing and short-lived local caches.

Study: Instagram

Learn how Instagram is built from the people who explain it. Watch first for the picture, then read for the detail.

0 of 5 watched

24 min

Start here

Designing INSTAGRAM: System Design of News Feed

Gaurav Sen · 763K views

Then

Cookie preferences