Design Instagram: media upload, CDN, and feed ranking
Instagram's Feed Service: Photos to 2B Users
Adding the Feed Service closes the loop: when @cristiano posts, Instagram fans that post out to 600M followers' timeline caches. When you open your feed, CDN URLs in your timeline resolve directly to edge-cached photos in milliseconds.
Request Flow
Photo upload triggers fanout
After upload completes and metadata is saved to the Database, the API calls the Feed Service to fanout the new post to all followers' timeline caches.
feedService.fanout({
authorId: userId,
photoId,
cdnUrl: 'https://cdn.example.com/photos/...',
timestamp: now
})Feed Service builds timelines
The Feed Service pushes the post to the timeline cache of each follower. For normal users (< 1M followers), this is a write-time fanout — reads become O(1) cache reads.
Viewer reads cached timeline
The Viewer's GET /feed returns a list of CDN URLs. No S3 reads. No heavy DB joins. The timeline is a pre-assembled list of content identifiers pointing to edge-cached images.
GET /feed
→ [
{ url: 'cdn.example.com/photo1_md.jpg', ... },
{ url: 'cdn.example.com/photo2_md.jpg', ... }
]Why We Need This
Without fanout, every viewer's timeline requires joining follows → posts → photo metadata at read time — an O(followers × recent_posts) query that doesn't scale past thousands of users.
Key Insight
The feed service turns expensive read-time fan-in into cheap write-time fan-out. Combined with CDN image delivery, a viewer's entire feed can be served from cache with zero origin reads.
Step 6 of 7
