Build a scalable Content Delivery Pipeline

Question

Design a scalable content delivery system that handles millions of requests. Discuss trade-offs in consistency, availability, and performance.

Codemia · Accepted Answer

### Functional Requirements:
1. **User Content Delivery**: The system must deliver video content to users with minimal latency.
2. **Content Management**: Ability to upload, process, and store videos with metadata tagging.
3. **User Interaction**: Support for likes, comments, and shares on content.
4. **Rate Limiting**: Implement API rate limiting to ensure fair use across users.

### Non-Functional Requirements:
1. **Scalability**: Must scale horizontally to handle millions of concurrent requests.
2. **Availability**: Aim for 99.99% uptime with quick recovery from failures.
3. **Performance**: Average latency under 200ms for video delivery.
4. **Monitoring and Logging**: Real-time monitoring with alerting for system anomalies.

Assuming TikTok has approximately 1 billion active users, with an average of 5 video requests per user per day:
- **Daily Requests**: 1 billion users * 5 requests/user = 5 billion requests/day.
- **Requests per Second (RPS)**: 5 billion requests/day ÷ 86400 seconds/day = ~57,870 RPS.
- **Content Size**: Average video size is around 10MB. 
- **Total Data per Day**: 5 billion requests * 10MB/request = 50 petabytes/day. 
This indicates the need for a highly efficient caching mechanism and distributed storage to manage this scale.

### Component Diagram:
- **API Gateway**: Handles incoming requests and routes them to appropriate services.
- **Load Balancer**: Distributes incoming requests across multiple instances of the content delivery service.
- **Content Delivery Network (CDN)**: Offloads traffic and caches video content at edge locations globally.
- **Video Processing Service**: Handles encoding, decoding, and metadata extraction upon upload.
- **Database**: Use **Cassandra** for metadata storage, providing horizontal scalability and high availability.
- **Message Queue**: Implement **Kafka** for asynchronous processing of upload events and notifications.

### Technology Choices:
- **CDN**: Cloudflare or Akamai for fast global content delivery.
- **Load Balancer**: NGINX or AWS ELB for distributing traffic.

### Schema Design:
- **User Table**:
  - `user_id`: UUID (Primary Key)
  - `username`: String
  - `email`: String

- **Video Table**:
  - `video_id`: UUID (Primary Key)
  - `user_id`: UUID (Foreign Key)
  - `url`: String (Location in the CDN)
  - `metadata`: JSON (Contains likes, comments, etc.)
  - `created_at`: Timestamp

- **Interactions Table**:
  - `interaction_id`: UUID (Primary Key)
  - `video_id`: UUID (Foreign Key)
  - `user_id`: UUID (Foreign Key)
  - `type`: Enum (like, comment, share)
  - `created_at`: Timestamp

This model supports quick access to video metadata and user interactions.

### Architectural Decisions:
1. **Consistency vs. Availability**: Chose eventual consistency for user interactions. This means that while video metadata may not be instantly up-to-date across all replicas, the system will remain available to handle user queries and uploads.
2. **Caching Strategy**: Implemented an aggressive caching strategy using Redis to cache frequently accessed videos, trading off some consistency for improved performance and reduced database load.
3. **Asynchronous Processing**: By using Kafka for video processing, we ensure that the upload pipeline is decoupled from delivery, allowing users to upload new content without waiting for processing to complete, thus enhancing user experience.
4. **Rate Limiting**: Implemented token bucket algorithm at the API gateway for fair usage, ensuring no single user can overwhelm the system.

Build a scalable Content Delivery Pipeline

TikTok

What the Interviewer Expects

Key Topics to Cover

How to Approach This

Possible Follow-up Questions

Practice a Similar Problem on Codemia

Sample Answer

Requirements

Functional Requirements:

Capacity Estimation

Submit Your Answer

TikTok Software Engineer Interview Guide

Related Questions

Design a high-throughput Inventory Management System

Design a low-latency Rate Limiting System

Design a fault-tolerant Payment System

Design a fault-tolerant Messaging System

Design Walmart Product Search