Social Media (Instagram)

Overview

Designing a Social Media platform (Instagram) combines photo/video storage, news feed, real-time features, and discovery. The question tests your ability to handle media uploads at scale, feed generation (push vs pull), stories (ephemeral content), and explore/recommendations. This design matters in interviews because it's a well-known product with multiple subsystems—each with its own scaling challenges. Demonstrating you understand CDN for media, fan-out for feed, and how to design for both read-heavy (feed) and write-heavy (upload) workloads shows you can architect complex social products. Meta and other companies ask this to see if you can reason about the full stack from client to storage.

Requirements

Functional

Upload photos/videos with captions
Follow users, view feed of posts
Stories (24-hour ephemeral)
Like, comment, direct message
Explore/discover content
Search users and hashtags

Non-Functional

Low latency — feed load <200ms
High availability — 99.99%
Media delivery — CDN, multiple resolutions
Scalability — billions of posts, millions of uploads/day

Capacity Estimation

Assume 1B users, 100M posts/day. 50M stories/day. 10B feed reads/day. Media: 100M * 5MB = 500TB/day upload.

Architecture Diagram

Component Deep Dive

Media Upload Service

Accepts uploads, stores in object store. Generates thumbnails, multiple resolutions. Async processing.

Feed Service

Generates feed from followed users. Push (pre-compute) or pull. Ranking by engagement.

Story Service

Ephemeral content, 24h TTL. Separate pipeline. High write, high read, auto-delete.

CDN

Serves images/videos. Multiple resolutions. Reduces origin load.

Search/Discovery

Elasticsearch for users, hashtags. Recommendation engine for explore.

Notification Service

Push for likes, comments, new followers. Real-time delivery.

Database Design

Posts, users, follows in PostgreSQL/Cassandra. Media URLs in object store. Feed in Cassandra (push) or computed (pull). Stories in Redis with TTL or Cassandra with TTL.

API Design

Method	Path	Description
`POST`	/api/posts	Upload post. Multipart. Returns post_id.
`GET`	/api/feed	Get feed. Paginated.
`POST`	/api/stories	Post story. Expires in 24h.
`GET`	/api/explore	Discover content. Personalized.

Scalability & Trade-offs

Feed: Push for most users; pull for celebrities (avoid write amplification).
Media: Multiple resolutions increase storage; single resolution + on-demand transcoding reduces storage, adds latency.
Stories: Separate storage with TTL vs reusing post table. TTL simplifies cleanup.

Related System Designs

Social