Designing a Photo Sharing App

A photo sharing platform lets users upload, process, and share images with followers.

Requirements

• Upload photos up to 10MB
• Apply filters
• Follow users, view feed
• Explore / discover content
• Hashtags, location tags
• 500M users, 100M photos/day

Photo Upload Pipeline

Client uploads photo
    ↓
Upload Service → S3 (original)
    ↓
Message Queue (Kafka)
    ↓
Image Processing Workers:
  - Resize: thumbnail, medium, large
  - Apply filter (optional)
  - Extract EXIF data
  - Generate blurhash (placeholder)
    ↓
S3 (processed variants)
    ↓
CDN distribution
    ↓
Update photo metadata in DB

Storage Estimation

100M photos/day × 3 variants × 200KB avg = 60TB/day
→ CDN caches hot content
→ S3 Intelligent Tiering for cold storage

Metadata Database

photos(id, user_id, caption, location, created_at)
photo_urls(photo_id, size, s3_url)
hashtags(id, tag)
photo_hashtags(photo_id, hashtag_id)

Feed Generation

Same as news feed:

• Hybrid fan-out
• Redis sorted set per user (score = timestamp)
• Celebrity posts merged at read time

Explore / Discover

Content signals → ML pipeline → Explore feed
  - Post engagement rate
  - Similarity to liked content
  - Geographic proximity
  - Trending hashtags

Hashtag Indexing

Post created → Extract hashtags
            → Index in Elasticsearch

Hashtag search: #sunset
→ Elasticsearch query → Recent + popular photos

Stories (24-hour ephemeral)

Story uploaded → S3
Story metadata → Redis (TTL: 24 hours)
Viewer list   → Redis SET per story
→ Auto-expire after 24 hours

Conclusion

Photo apps are CDN-first. Async image processing, hybrid feed fan-out, and Elasticsearch-powered hashtag search form the core architecture.