Discord WhatsApp

Scaling Hotstar to 25 million concurrent users

· Tech/Architecture · 2 min read · #scaling

How Hotstar Application Scaled 25 Million Concurrent Users | Performance Testing | Load Testing

https://www.youtube.com/watch?v=9b7HNzBB3OQ

Points touched upon

  • How Hotstar handled 25.3 million concurrent viewers during the India vs New Zealand Cricket World Cup match, setting a global live-streaming record
  • The challenges of sudden traffic spikes exceeding one million users per minute, and how viewers leaving a live stream can overwhelm homepage and backend services.
  • Why traditional AWS Auto Scaling was insufficient, including slow instance provisioning, capacity shortages, API throttling and availability zone limitations.
  • Hotstar's approach to proactive scaling, maintaining spare capacity, using preconfigured machine images and developing custom scaling tools based on traffic and concurrency.
  • The use of multiple load balancers, weighted routing, different EC2 instance types and distributed infrastructure to overcome scaling bottlenecks.
  • Project Hulk and large-scale load testing using thousands of EC2 instances across eight AWS regions to simulate traffic from up to 50 million concurrent users.
  • The role of chaos engineering in testing network failures, increased latency, delayed scaling, sudden traffic changes and failures across dependent services.
  • The importance of identifying system breaking points, understanding user journeys, prioritizing essential features and preparing infrastructure for failures before major live events.
Non-profit informational compilation. Views are not the compiler’s; no liability assumed. Notes and structure are human-authored; some outlines may be AI-assisted for grammar and clarity.