Youtube Design

Problems

Storage?

Scalability?

Web Server?

Cache?

Web Server

* Multiple Web Servers behind load balancer

* User Authentication

* Sessions

* Fetching and Updating Users Data

* Video Server

* Multiple Web Servers behind load balancer

* Recommendation Server

* Multiple Web Servers behind load balancer

Storage

MySQL Database:

Lets assume videos have comments and likes and comments do not have likes

Table:UserDetails

Fields: UserId, Name, Email, Reg Data, Profile Information(Address, Age etc.)

Table: User

Fields: UserId, Email, Username, Password, Reg Date

Table: Video

Fields: VideoId, UserId, Title, Description, Size, Video File, LikeCount, ViewCount

Table: VideoComments

Fields: VideoId, UserId, Comment, Date

Table: VideoLike

Fields: VideoId, UserId, Date

SQL is Geo Replicated

Video and Image Storage

Thumbnails of different sizes for different screens.

CDN - Proxy servers deployed in multiple data centres. Provide Availability, Scalability and Performance

Popular videos go on CDN. less popular ones go on local storage

Database Scaling

Simple DB

Master/Slave

Partitioning based on location. So requests can be served quickly (Sharding approaches)

Video Cluster is allocated the max resources. General cluster is the one with basic resources. Bottleneck - watching videos.

Cache

Front and back end cache. Not on videos as there are many many videos.

Security

Problems:

• Hack View Count
• Programmatically send view count

Remedy: Smart logic implementation

• IP Address block
• Cap on view count on the IP Address
• Also can observe browser agent and user's past history
• View count also relates to comments, share count, view time etc.