Requirements

Functional Requirements:

• Count views for each video accurately, ensuring every legitimate view is counted, even under high traffic.
• Display real-time or near-real-time view counts to users and content creators.
• Prevent fraudulent or bot views using techniques like watch duration thresholds, rate limiting, IP analysis, CAPTCHA, or device fingerprinting.
• Provide content creators with view count trends broken down by time period (daily, weekly, monthly) for analytics and insights.

Non-Functional Requirements:

• Scalability
• Reliability & Availability
• Performance
• Security
• Fault Tolerance
• Consistency

API Design

Assuming each video has a unique ID and a known length in seconds.

• videoWatchStarted(videoID, startTimestamp)
• videoWatchEnded(videoID, endTimestamp)
• videoViewCount(videoID, periodFrom, periodTo): Number
• publishNotification(videoID, timestamp)

High-Level Design

• The Load Balancer is to ensure Scalability, Reliability & Availability NFRs are guaranteed by distributing load among horizontally scalable application servers.
• The Geo DNS is to ensure Fault Tolerance NFR is guaranteed by failing over to a different region in case of failure of one of the available regions.
• Application Server
  • The Authentication Service is to guarantee security to prevent fraudulent or bot views and also to identify the person who watched a specific video for analytics purposes.
  • View Count Service is the service responsible for counting video views according to specific rules.
  • Analytics Service is the service that will be used by the client to query analytics data. For example, view count trends for a specific period of time.
  • Notification Workers is the service responsible for notifying content creators once a video view has been added to the database.
• The Redis server is to guarantee consistency NFR across application servers when it comes to counting views.
• The Relational Database is to guarantee persistence of the view data. Also it's relational since aggregation queries to count views will be used for analytics.
• Message Queue allows the View Count Service to publish video view notifications to a queue/topic and it also allows Notification Workers to get notified once a video view notification is published and consequently notifies content creators either through email or WebSockets.

Detailed Component Design

• videoWatchStarted(videoID, startTimestamp)
  • this function will store the video id and timestamp in a session variable
• videoWatchEnded(videoID, endTimestamp)
  • this function will compare the the length watched by comparing the watched length (endTimestamp - startTimestamp) in seconds to the length of the video.
  • If the length watched is at least 70% of the video, add a timestamp for this view.
    • We can also store some extra information like who watched it, where it was watched, and may be collect some information about the browser and operating system used to watch it.
• videoViewCount(videoID, periodFrom, periodTo): Number
  • this function will query the database to obtain the number of view timestamps between periodFrom and periodTo of this videoID and return a number.
• publishNotification(videoID, timestamp)
  • This function will be used by View Count Service to publish a message to the queue/topic once a video view event occurs.