Design Facebook Live Video Streaming

Question

Design a live video streaming system at Facebook's scale. Cover video ingestion, transcoding, CDN distribution, adaptive bitrate streaming, and handling millions of concurrent viewers on a single stream.

Codemia · Accepted Answer

### Functional Requirements
1. **Live Video Ingestion**: Allow users to stream live video content from various devices (smartphones, webcams, etc.).
2. **Transcoding**: Convert incoming video streams into multiple formats and resolutions for adaptive bitrate streaming.
3. **CDN Distribution**: Distribute the video streams through a Content Delivery Network (CDN) to ensure low latency and high availability.
4. **Adaptive Bitrate Streaming**: Automatically adjust video quality based on viewer's network conditions.
5. **Viewer Interaction**: Enable real-time interactions such as comments, reactions, and polls during the live stream.

### Non-Functional Requirements
1. **Scalability**: Support millions of concurrent viewers and scale horizontally as needed.
2. **Latency**: Ensure low latency (under 2 seconds) for live streaming.
3. **Reliability**: Achieve 99.99% uptime for live streaming services.
4. **Security**: Protect against unauthorized access and ensure the integrity of video content.
5. **Analytics**: Provide real-time analytics on viewer engagement and stream quality.

Assuming we want to support 10 million concurrent viewers for a single stream:
- **Bitrate Calculation**: Average bitrate for HD live streaming ~4 Mbps.
- **Total Bandwidth Requirement**: 10 million viewers * 4 Mbps = 40,000 Gbps.
- **Transcoding Needs**: Each stream will be transcoded into 3 resolutions (720p, 480p, 360p). For each resolution, the average bitrate would be:
  - 720p: 4 Mbps
  - 480p: 2 Mbps
  - 360p: 1 Mbps
- **Concurrent Transcoding**: Assuming a maximum of 100 concurrent streams, bandwidth after transcoding will be:
   - 100 streams * (4 + 2 + 1) Mbps = 700 Mbps.
- **Storage Requirements**: Assuming 30 minutes of buffer for replays:
   - 30 minutes * 10 million viewers * average bitrate (4 Mbps) = 600 TB of storage per stream.

### Component Diagram
- **Video Ingestion Service**: Handles incoming streams using RTMP or WebRTC.
- **Transcoding Service**: Uses FFmpeg or AWS Elemental for transcoding streams into multiple resolutions.
- **Storage Layer**: Uses Amazon S3 for storing live video segments and metadata.
- **CDN**: Leverage Facebook's own CDN or integrate with Akamai/Cloudflare for global distribution.
- **Viewer Service**: Manages connections to viewers, providing adaptive bitrate streaming using HLS/DASH protocols.
- **Real-time Interaction Service**: Handles comments and reactions using WebSockets for low-latency interaction.
- **Analytics Service**: Collects data on viewer engagement and stream performance using Apache Kafka and stores in a time-series database like InfluxDB.

### Schema Design
#### Stream Metadata Table
| Field            | Type         | Description                          |
|------------------|--------------|--------------------------------------|
| stream_id        | UUID         | Unique identifier for the stream     |
| user_id          | UUID         | ID of the user streaming              |
| title            | String       | Title of the stream                  |
| start_time       | Timestamp    | Time when the stream started         |
| end_time         | Timestamp    | Time when the stream ended           |
| status           | Enum         | Status (live, ended, error)         |

1. **Transcoding Delay vs. Quality**: Choosing to transcode in real-time can introduce latency. We need to balance between high-quality streams and acceptable delay by using pre-defined profiles and fast transcoding hardware.
2. **Cost vs. Performance**: Using high-end servers for transcoding and storage can improve performance but increase costs. We must consider using spot instances or auto-scaling strategies to manage costs effectively.
3. **Data Consistency vs. Availability**: In a distributed system, ensuring real-time analytics while maintaining high availability can be challenging. We may opt for eventual consistency in analytics data to favor lower latency.
4. **CDN Choice**: Using Facebook’s own CDN can offer better integration and performance, but reliance on third-party CDNs may provide more global reach and redundancy.

Design Facebook Live Video Streaming

Meta

How to Approach This

Sharpen Your Skills on Codemia

Sample Answer

Requirements

Functional Requirements

Capacity Estimation

Submit Your Answer

Meta 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