Design a Video Streaming for NVIDIA

Question

Design a multi-tenant video streaming system that handles millions of requests. Discuss trade-offs in consistency, availability, and performance.

Codemia · Accepted Answer

### Functional Requirements
1. **Multi-Tenant Support**: The system must handle multiple tenants (different organizations) simultaneously without data leakage.
2. **Video Upload/Processing**: Users can upload videos, which are processed (encoded, transcoded).
3. **Video Playback**: Users can stream videos with adaptive bitrate streaming to accommodate varying network conditions.
4. **User Authentication**: Support secure user authentication and authorization for accessing video content.
5. **Analytics Dashboard**: Provide usage statistics for tenants, including view counts and performance metrics.

### Non-Functional Requirements
1. **Scalability**: Must handle millions of concurrent users and requests with low latency.
2. **Availability**: Aim for 99.99% uptime, ensuring that the service is always available.
3. **Performance**: Video start times should be under 2 seconds and should support high-definition streaming.
4. **Security**: Ensure encryption of video data in transit and at rest, along with strict access controls.

### Back-of-Envelope Calculations
1. **User Base**: Assume 10 million active users.
2. **Concurrent Users**: Assume 10% concurrent usage at peak, leading to 1 million concurrent users.
3. **Requests per Second (RPS)**: If each user makes 10 requests (e.g., play, pause) per minute, we have:
   - RPS = (1,000,000 users * 10 requests) / 60 seconds = ~166,667 RPS.
4. **Storage**: Assume average video size of 1GB, with 100,000 videos:
   - Total Storage = 100,000 videos * 1GB = 100TB.
5. **Bandwidth**: For HD streaming at ~5 Mbps, during peak:
   - Bandwidth = 1,000,000 users * 5 Mbps = ~5 Tbps.

### Component Diagram
- **Frontend**: A web application and mobile app for users to upload and stream videos.
- **Load Balancer**: Distributes incoming requests across multiple application servers.
- **Application Servers**: Handle video upload, processing, and streaming logic. Technologies: Node.js with Express for handling requests.
- **Video Processing Service**: Utilizes NVIDIA GPUs for transcoding and encoding videos, using FFmpeg.
- **Storage**: Use Amazon S3 for object storage of videos, and DynamoDB for metadata storage.
- **CDN**: CloudFront or Akamai for distributing video content globally.
- **Database**: NoSQL database (DynamoDB) for user data and video metadata.
- **Analytics Service**: Processes logs and provides usage statistics, implemented with Apache Kafka and Spark.

### Schema Design
1. **Users Table (DynamoDB)**:
   - `UserID (PK)`: String
   - `TenantID`: String
   - `Email`: String
   - `PasswordHash`: String
   - `CreatedAt`: Timestamp

2. **Videos Table (DynamoDB)**:
   - `VideoID (PK)`: String
   - `UserID`: String (FK)
   - `TenantID`: String
   - `VideoURL`: String
   - `Metadata`: JSON (e.g., title, description, duration)
   - `CreatedAt`: Timestamp

3. **Analytics Table (DynamoDB)**:
   - `VideoID (PK)`: String
   - `UserID`: String
   - `ViewsCount`: Number
   - `LastViewed`: Timestamp

### Access Patterns
- Fast retrieval of user data for authentication and video metadata.
- Efficient querying for analytics data.

### Architectural Decisions
1. **Consistency vs. Availability**: Choosing eventual consistency for video metadata in DynamoDB to prioritize availability and performance. This is acceptable as slight delays in reflecting changes (like new uploads) are tolerable for user experience.
2. **Caching**: Using Redis for caching frequently accessed video metadata to reduce load on the database and improve performance.
3. **Replication**: Multi-region replication for DynamoDB and S3 to ensure low-latency access and high availability, but introduces complexity in ensuring data consistency across regions.
4. **Security**: Implementing OAuth 2.0 for user authentication adds complexity but enhances security. Data encryption in transit and at rest ensures compliance with security requirements, but may impact performance slightly.
5. **Load Balancing**: Choosing a Layer 7 load balancer to make routing decisions based on the content type (e.g., video vs. metadata requests) for optimized resource usage.

Design a Video Streaming for NVIDIA

NVIDIA

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

Back-of-Envelope Calculations

Submit Your Answer

NVIDIA 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