Practice/Meta/Design Instagram

Design Instagram

Product DesignMust

Problem Statement

Design a large-scale video streaming service similar to YouTube or Netflix where users can upload videos, watch content, and receive personalized recommendations. The platform needs to handle millions of concurrent viewers streaming high-definition video content while maintaining smooth playback with minimal buffering. Content creators should be able to upload videos of varying lengths and formats, which the system must process, encode, and distribute efficiently across a global user base.

The system must support features like video search, comments, likes, subscriptions, and viewing history. Consider how to optimize for both live streaming and on-demand video playback, handle content delivery at scale, and ensure videos load quickly regardless of user location or network conditions.

Key Requirements

Functional

Video Upload and Processing -- Users can upload videos up to 2GB in size; system must transcode into multiple resolutions and formats
Video Streaming -- Users can watch videos with adaptive bitrate streaming based on network conditions
Search and Discovery -- Users can search for videos by title, tags, or content; platform provides recommendations
Social Features -- Users can like, comment, subscribe to channels, and share videos
Analytics Dashboard -- Content creators can view watch time, viewer demographics, and engagement metrics

Non-Functional

Scalability -- Support 100 million daily active users with 500,000 concurrent video streams; handle 10,000 video uploads per hour
Reliability -- 99.9% uptime for streaming service; no data loss for uploaded videos
Latency -- Video playback should start within 2 seconds; seek operations complete in under 1 second
Consistency -- View counts and likes can be eventually consistent; user authentication must be strongly consistent

What Interviewers Focus On

Based on real interview experiences, these are the areas interviewers probe most deeply:

1. Video Storage and Content Delivery Architecture

This tests your understanding of how to efficiently store and serve large binary files at global scale. Raw video files are enormous and serving them directly from origin servers would be prohibitively expensive and slow.

Hints to consider:

Separate storage strategy for raw uploads versus processed/encoded videos
Content Delivery Network (CDN) architecture for edge caching and geographic distribution
Chunk-based storage to enable seeking and adaptive bitrate streaming
Cost tradeoffs between hot storage (frequently accessed) and cold storage (archival)

2. Video Processing Pipeline

Interviewers want to see how you design asynchronous job processing at scale. Video transcoding is CPU-intensive and time-consuming, requiring careful orchestration.

Hints to consider:

Message queue-based architecture to decouple upload from processing
Parallel processing of different resolution outputs (360p, 720p, 1080p, 4K)
Thumbnail generation and preview clip extraction as separate pipeline stages
Handling failures and retries when encoding jobs crash or timeout

3. Adaptive Bitrate Streaming Protocol

This reveals your knowledge of video streaming technology and how to optimize user experience across varying network conditions.

Hints to consider:

HTTP Live Streaming (HLS) or MPEG-DASH protocol selection and tradeoffs
Manifest files that describe available quality levels and segment locations
Client-side logic for detecting bandwidth and switching between quality tiers
Segment duration choices balancing latency versus overhead

4. Database Schema and Metadata Management

Video metadata (title, description, views, likes) has different access patterns than the video content itself, requiring thoughtful data modeling.

Hints to consider:

Separate databases for user data, video metadata, and social interactions
Denormalization strategies for read-heavy operations like view counts
Indexing strategy for search functionality across millions of videos
Time-series data storage for analytics and watch history

5. Recommendation System Architecture

This explores your ability to design personalized experiences using both real-time and batch processing patterns.

Hints to consider:

Collaborative filtering based on viewing history of similar users
Content-based filtering using video tags, categories, and metadata
Real-time signal incorporation (clicks, watch duration, engagement) versus batch model training
Cold start problem for new users and new videos without historical data

Practice/Meta/Design Instagram

Design Instagram

Product DesignMust

Problem Statement

Key Requirements

Functional

Video Upload and Processing -- Users can upload videos up to 2GB in size; system must transcode into multiple resolutions and formats
Video Streaming -- Users can watch videos with adaptive bitrate streaming based on network conditions
Search and Discovery -- Users can search for videos by title, tags, or content; platform provides recommendations
Social Features -- Users can like, comment, subscribe to channels, and share videos
Analytics Dashboard -- Content creators can view watch time, viewer demographics, and engagement metrics

Non-Functional

Scalability -- Support 100 million daily active users with 500,000 concurrent video streams; handle 10,000 video uploads per hour
Reliability -- 99.9% uptime for streaming service; no data loss for uploaded videos
Latency -- Video playback should start within 2 seconds; seek operations complete in under 1 second
Consistency -- View counts and likes can be eventually consistent; user authentication must be strongly consistent

What Interviewers Focus On

Based on real interview experiences, these are the areas interviewers probe most deeply:

1. Video Storage and Content Delivery Architecture

Hints to consider:

Separate storage strategy for raw uploads versus processed/encoded videos
Content Delivery Network (CDN) architecture for edge caching and geographic distribution
Chunk-based storage to enable seeking and adaptive bitrate streaming
Cost tradeoffs between hot storage (frequently accessed) and cold storage (archival)

2. Video Processing Pipeline

Interviewers want to see how you design asynchronous job processing at scale. Video transcoding is CPU-intensive and time-consuming, requiring careful orchestration.

Hints to consider:

Message queue-based architecture to decouple upload from processing
Parallel processing of different resolution outputs (360p, 720p, 1080p, 4K)
Thumbnail generation and preview clip extraction as separate pipeline stages
Handling failures and retries when encoding jobs crash or timeout

3. Adaptive Bitrate Streaming Protocol

This reveals your knowledge of video streaming technology and how to optimize user experience across varying network conditions.

Hints to consider:

HTTP Live Streaming (HLS) or MPEG-DASH protocol selection and tradeoffs
Manifest files that describe available quality levels and segment locations
Client-side logic for detecting bandwidth and switching between quality tiers
Segment duration choices balancing latency versus overhead

4. Database Schema and Metadata Management

Video metadata (title, description, views, likes) has different access patterns than the video content itself, requiring thoughtful data modeling.

Hints to consider:

Separate databases for user data, video metadata, and social interactions
Denormalization strategies for read-heavy operations like view counts
Indexing strategy for search functionality across millions of videos
Time-series data storage for analytics and watch history

5. Recommendation System Architecture

This explores your ability to design personalized experiences using both real-time and batch processing patterns.

Hints to consider:

Collaborative filtering based on viewing history of similar users
Content-based filtering using video tags, categories, and metadata
Real-time signal incorporation (clicks, watch duration, engagement) versus batch model training
Cold start problem for new users and new videos without historical data