Design YouTube (used)

December 5, 2021 7:00 PM PST

---

This document summarizes a mock system design interview focused on designing a video-sharing platform similar to YouTube. The interview covered functional and non-functional requirements, system design, and various technical discussions related to video upload, storage, and retrieval.

Interview Details

• Level: L5 (Senior)
• Duration: 1 hour
• Drawing Tool Used: Whimsical

Requirements

Functional Requirements

• View videos and thumbnails
• Upload videos
• Search for videos
• Display popularity and comments on videos

Non-Functional Requirements

• Ensure video integrity (no loss of videos)
• Provide smooth video playback
• Ensure high availability
• Minimize network costs

Constraints

• Daily Active Users (DAU): 100 million
• Average uploads: 5 videos per person per day
• Queries per second (QPS) for views: ~5000
• Upload and download ratios and bandwidth requirements were discussed.

System Design

External APIs

1. Video Retrieval

   • getVideo(video_id, user_id, offset)
     • Offset is necessary due to large video sizes being divided into smaller segments.

2. Video Upload

   • uploadVideo(video_id, user_id, description, length, tags[], video_content)
     • Returns a presigned URL for uploading.

Upload Flow

• The upload process involves changing the API to accommodate presigned URLs.
• Discussion on whether to use Google Cloud Storage or AWS S3 versus maintaining a custom distributed file system.
• Consideration of trade-offs between pre-built solutions and self-maintained systems.

Encoding and Processing

• A message queue is introduced to trigger the encoding service, addressing the time-consuming nature of encoding and uploading.
• Discussion on the choice of storage solutions, emphasizing the need for scalable storage like S3.

Download Flow

• Data retrieval from S3 with the addition of a CDN to cache popular videos.
• Implementation of a TopK system to identify frequently watched videos for caching.
• Optimization strategies include encoding videos at different resolutions.

Comment Features

• APIs for commenting and liking were proposed:
  • comment(video_id, user_id, content, comment_id)
  • like(video_id, comment_id)
• Discussion on database schema and Redis usage for caching comments and handling failures.

Database Schema

• Considerations for handling Redis failures and ensuring data integrity through synchronization with NoSQL databases.
• Discussion on the importance of maintaining a backup Redis instance.

Additional Design Considerations

• Importance of requirement gathering and feature prioritization during the interview.
• The interviewee should drive the discussion and confirm understanding with the interviewer.
• Discussion on handling upload errors, pacing of uploads, and the scalability of the system.
• Consideration of various storage solutions and their implications on performance and cost.

Audience Feedback

• Emphasis on the need for the interviewee to propose ideas and follow up with questions.
• Discussion on the importance of covering all features within the time constraints of the interview.
• Suggestions for improving the systematic expression of knowledge during the interview.

Technical Discussions

• Various technical aspects were discussed, including:
  • Handling of uploads and downloads, including chunking and encoding.
  • Use of message queues for asynchronous processing.
  • Considerations for using NoSQL versus SQL databases.
  • Strategies for caching and ensuring high availability.

This summary encapsulates the key points discussed during the mock interview, highlighting the technical depth and considerations necessary for designing a scalable video-sharing platform.