Why video streaming handles massive data in HLD - Design It to Understand It

Design: Video Streaming Data Handling

Focus on data handling aspects of video streaming including storage, delivery, and scaling. Exclude detailed UI design and content creation.

Functional Requirements

FR1: Support streaming of video content to millions of users simultaneously

FR2: Deliver video with minimal buffering and latency

FR3: Handle different video qualities and formats

FR4: Allow users to pause, rewind, and fast-forward videos

FR5: Support live streaming and on-demand videos

Non-Functional Requirements

NFR1: Scale to handle millions of concurrent viewers

NFR2: Maintain p99 latency under 300ms for video start time

NFR3: Ensure 99.9% availability for streaming service

NFR4: Efficiently store and serve large video files

NFR5: Optimize bandwidth usage to reduce costs

Think Before You Design

Questions to Ask

❓ Question 1

❓ Question 2

❓ Question 3

❓ Question 4

❓ Question 5

Key Components

Video storage system (object storage or distributed file system)

Content Delivery Network (CDN)

Video encoding and transcoding services

Load balancers and streaming servers

Caching layers and edge servers

Design Patterns

Content Delivery Network (CDN) for caching and distribution

Adaptive Bitrate Streaming to adjust video quality dynamically

Sharding and partitioning of video data

Asynchronous processing for encoding and uploading

Load balancing for streaming servers

Reference Architecture

User Devices
   |
   v
Load Balancer
   |
Streaming Servers <--> Video Encoding Service
   |
CDN Edge Servers
   |
Distributed Video Storage

Components

Load Balancer

Nginx or AWS ELB

Distributes user requests evenly to streaming servers

Streaming Servers

Custom streaming server or Wowza

Handles video streaming sessions and user controls

Video Encoding Service

FFmpeg or cloud encoding services

Transcodes videos into multiple formats and bitrates

Content Delivery Network (CDN)

Cloudflare, Akamai, AWS CloudFront

Caches video content close to users to reduce latency

Distributed Video Storage

Amazon S3, Google Cloud Storage, or HDFS

Stores original and transcoded video files reliably

Request Flow

1. User requests video from device.

2. Request hits Load Balancer which routes to Streaming Server.

3. Streaming Server checks CDN cache for requested video segment.

4. If cache miss, Streaming Server fetches video segment from Distributed Video Storage.

5. Video segment delivered to CDN edge server and then to user device.

6. Video Encoding Service processes uploaded videos into multiple qualities asynchronously.

7. CDN caches popular video segments to serve future requests quickly.

Database Schema

Entities: - Video: id, title, description, upload_date, duration - VideoSegment: id, video_id (FK), quality, format, storage_path - User: id, username, subscription_type - StreamingSession: id, user_id (FK), video_id (FK), start_time, end_time Relationships: - One Video has many VideoSegments - One User can have many StreamingSessions - StreamingSessions link Users to Videos watched

Scaling Discussion

Bottlenecks

Storage capacity and throughput for large video files

Network bandwidth for streaming to many users

Encoding service processing time for new videos

Load on streaming servers during peak usage

Cache misses causing higher latency

Solutions

Use scalable object storage with automatic replication and partitioning

Deploy CDN with global edge locations to reduce bandwidth load on origin

Use distributed and parallel encoding pipelines

Auto-scale streaming servers based on traffic

Pre-warm CDN caches and use predictive caching algorithms

Interview Tips

Time: 10 minutes to clarify requirements and constraints, 15 minutes to design architecture and data flow, 10 minutes to discuss scaling and bottlenecks, 10 minutes for Q&A

Explain why video data is massive due to size and concurrent users

Discuss importance of CDN to reduce latency and bandwidth

Highlight adaptive bitrate streaming for user experience

Mention asynchronous encoding to handle large uploads

Describe scaling strategies for storage, network, and compute