📝 Problem Description
Design a video streaming platform like YouTube or Netflix that allows users to upload, process, and stream videos. The system should support adaptive bitrate streaming, global content delivery, recommendations, and handle billions of daily video views with minimal buffering.
👤 Use Cases
1.
Creator wants to upload a video so that the video is processed and made available for streaming
2.
Viewer wants to search for videos so that they find relevant content
3.
Viewer wants to play a video so that they watch with minimal buffering
4.
Viewer wants to switch video quality so that stream adapts to their bandwidth
5.
Viewer wants to like, comment, subscribe so that their engagement is recorded
6.
System wants to recommend videos so that personalized content for each user
✅ Functional Requirements
- •Upload videos (up to 12 hours, various formats)
- •Transcode videos to multiple resolutions (144p to 4K)
- •Stream videos with adaptive bitrate (ABR)
- •Search videos by title, description, tags
- •Recommend personalized content
- •Support live streaming
- •Engagement features: likes, comments, subscriptions
- •Monetization: ads insertion
⚡ Non-Functional Requirements
- •Video start time (time to first byte) < 200ms
- •Minimal buffering (< 1% rebuffer rate)
- •Support 5 billion video views per day
- •500 hours of video uploaded per minute
- •Global availability with low latency
- •High durability - never lose uploaded content
⚠️ Constraints & Assumptions
- •Video upload size: up to 256GB
- •Maximum video duration: 12 hours
- •Transcode to 6+ quality levels (144p, 240p, 360p, 480p, 720p, 1080p, 4K)
- •Thumbnail extraction at multiple timestamps
- •Global CDN coverage in 100+ countries
📊 Capacity Estimation
👥 Users
2B monthly active users, 1B daily views
💾 Storage
1 exabyte total, 500PB/year new content
⚡ QPS
Video plays: 50K/sec, Uploads: 1K/min
🌐 Bandwidth
100+ Tbps global egress
📐 Assumptions
- • 5B video views per day
- • 500 hours uploaded per minute = 720K hours/day
- • Average video: 10 minutes, 500MB after encoding
- • 80% of traffic goes to 20% of videos (hot content)
- • Average viewer watches 40 minutes/day
- • 6 encoding profiles per video
💡 Key Concepts
CRITICAL
Adaptive Bitrate Streaming
HLS/DASH protocols allow clients to switch quality based on network conditions. Video is split into small segments (2-10s) at multiple quality levels.
CRITICAL
CDN Edge Caching
90%+ of video requests served from edge cache. Popular content is pre-positioned at edge nodes. Cache hit rate directly impacts latency and cost.
CRITICAL
Video Transcoding
Convert source video to multiple profiles (codec, resolution, bitrate). Parallel processing for speed. FFmpeg is the standard tool.
HIGH
Chunked Upload
Large files uploaded in resumable chunks. Allows retry on failure, parallel upload. Direct to S3 with pre-signed URLs.
HIGH
Video Segmentation
Videos split into small segments (2-10s) for streaming. Enables seeking, quality switching, and efficient caching.
MEDIUM
Hot vs Cold Content
80% of views go to 20% of videos. Hot content cached aggressively. Cold content may stream from origin.
💡 Interview Tips
- 💡Separate upload/processing path from viewing path early
- 💡Explain adaptive bitrate streaming in detail - it's core to video
- 💡Discuss CDN extensively - it handles 90%+ of traffic
- 💡Mention transcoding pipeline parallelism
- 💡Cover the hot vs cold content distinction
- 💡Draw the video segment structure (HLS/DASH)