Transcoding and adaptive bitrate in HLD - System Design Exercise

Design: Video Streaming with Transcoding and Adaptive Bitrate

Design covers video ingestion, transcoding, storage, and adaptive bitrate delivery. Out of scope: DRM, user authentication, content recommendation.

Functional Requirements

FR1: Ingest live or on-demand video content from users or content providers

FR2: Transcode videos into multiple quality levels (bitrates and resolutions)

FR3: Store transcoded videos for on-demand playback

FR4: Deliver video streams to users with adaptive bitrate streaming based on their network conditions

FR5: Support at least 100,000 concurrent viewers

FR6: Ensure video startup latency under 3 seconds

FR7: Provide 99.9% availability for streaming service

Non-Functional Requirements

NFR1: Handle peak traffic of 100,000 concurrent viewers

NFR2: Latency for live streaming end-to-end should be under 10 seconds

NFR3: Storage must support petabytes of video data

NFR4: System should be scalable horizontally

NFR5: Use industry standard streaming protocols (HLS, DASH)

Think Before You Design

Questions to Ask

❓ Question 1

❓ Question 2

❓ Question 3

❓ Question 4

❓ Question 5

❓ Question 6

❓ Question 7

Key Components

Video ingestion service

Transcoding cluster

Storage system for original and transcoded videos

Content Delivery Network (CDN)

Streaming server or origin server

Client player with adaptive bitrate logic

Monitoring and alerting system

Design Patterns

Microservices architecture for modular components

Message queues for transcoding job management

Cache and CDN for low latency delivery

Adaptive bitrate streaming protocols (HLS, DASH)

Horizontal scaling of transcoding workers

Health checks and retries for fault tolerance

Reference Architecture

  +----------------+       +-------------------+       +----------------+
  | Video Ingestion| ----> | Transcoding Queue | ----> | Transcoding    |
  | Service       |       | (Message Queue)    |       | Workers        |
  +----------------+       +-------------------+       +----------------+
          |                                                        |
          v                                                        v
  +----------------+                                      +----------------+
  | Original Video |                                      | Transcoded     |
  | Storage        |                                      | Video Storage  |
  +----------------+                                      +----------------+
          |                                                        |
          v                                                        v
  +---------------------------------------------------------------+
  |                      Streaming Origin Server                  |
  +---------------------------------------------------------------+
                                  |
                                  v
                          +----------------+
                          | CDN (Edge Nodes)|
                          +----------------+
                                  |
                                  v
                          +----------------+
                          | Client Player  |
                          | (Adaptive Bitrate Logic) |
                          +----------------+

Components

Video Ingestion Service

HTTP/RTMP servers

Receive live or on-demand video uploads from content providers

Transcoding Queue

Message Queue (e.g., Kafka, RabbitMQ)

Manage and distribute transcoding jobs to workers asynchronously

Transcoding Workers

FFmpeg or cloud transcoding services

Convert original videos into multiple bitrates and resolutions

Original Video Storage

Object Storage (e.g., Amazon S3, MinIO)

Store raw uploaded videos reliably

Transcoded Video Storage

Object Storage with CDN integration

Store multiple quality versions of videos for streaming

Streaming Origin Server

HTTP server supporting HLS/DASH

Serve video manifests and segments to CDN

Content Delivery Network (CDN)

Commercial CDN or self-hosted edge caches

Deliver video content with low latency and high availability

Client Player

HTML5 video player with adaptive bitrate logic

Select appropriate video quality based on network conditions

Request Flow

1. 1. Content provider uploads video to Video Ingestion Service.

2. 2. Video Ingestion Service stores original video in Original Video Storage.

3. 3. Video Ingestion Service sends transcoding job message to Transcoding Queue.

4. 4. Transcoding Workers consume jobs, transcode video into multiple bitrates/resolutions.

5. 5. Transcoded videos are stored in Transcoded Video Storage.

6. 6. Streaming Origin Server generates manifests (HLS/DASH) referencing transcoded segments.

7. 7. CDN caches video manifests and segments from Streaming Origin Server.

8. 8. Client Player requests video manifest from CDN and starts playback.

9. 9. Client Player monitors network conditions and switches between bitrates adaptively.

Database Schema

Entities: - Video: id (PK), title, description, upload_time, uploader_id - TranscodingJob: id (PK), video_id (FK), status, created_at, completed_at - VideoSegment: id (PK), video_id (FK), bitrate, resolution, segment_url Relationships: - One Video has many TranscodingJobs - One Video has many VideoSegments representing different qualities

Scaling Discussion

Bottlenecks

Transcoding workers overwhelmed by high upload volume

Storage I/O bottleneck due to large video files

Streaming origin server unable to handle many manifest requests

CDN edge nodes overloaded during traffic spikes

Client network variability causing poor playback experience

Solutions

Scale transcoding workers horizontally; use autoscaling based on queue length

Use distributed object storage with high throughput and parallel access

Deploy multiple streaming origin servers behind load balancers

Leverage global commercial CDN with auto-scaling edge capacity

Implement robust adaptive bitrate algorithms and buffer management on client

Interview Tips

Time: Spend 10 minutes clarifying requirements and constraints, 20 minutes designing architecture and data flow, 10 minutes discussing scaling and trade-offs, 5 minutes summarizing.

Explain how transcoding enables multiple quality levels for diverse devices and networks

Describe asynchronous job processing for scalability and fault tolerance

Highlight use of CDN to reduce latency and improve availability

Discuss adaptive bitrate streaming benefits and client logic

Address scaling challenges and solutions realistically

Mention monitoring and failure handling strategies