Hotstar's journey of migrating infrastructure from EC2 to containers
summary
Hotstar migrates its services from EC2 instances to Kubernetes to handle peak loads — claims a live concurrency record of 25Mn+ viewers and notes spikes up to 1.1M RPS on specific services. The team replaces per-app EC2 fleets with containerized deployments, GoCD CI/CD, Terraform, Vault for secrets, Prometheus-driven HPAs for autopilot scaling, and an in-house cluster management layer. Outcomes: far fewer idle EC2 buffers, faster boot and standardized images, lower cost and faster rollouts; remaining pain points are node provisioning limits, image size and HPA tuning, and scaling cluster infra (Prometheus, logging).
AI-written summary · about
chapters (11)
- 0:00Introduction
- 0:42Scaling
- 1:44Infrastructure Platform
- 3:28Issues faced
- 10:00Kubernetes
- 13:52Challenges
- 17:43Load Testing
- 19:38IPL
- 21:02Comparison
- 23:29Application scaling
- 25:12Questions
related talks
29mAnalysing high throughput data in real-time
33mNavigating the Seas of Transformation: Migrating from EC2 to Kubernetes
22mKeynote: Kubernetes: As Seen On TV
38mTransforming Hulu’s Application Platform
32mMigrating 2 million CPU cores to Kubernetes
35mKubernetes journey
24mMigrating 2 million CPU cores to Kubernetes
26mLessons Learned Migrating an Existing Product to a Multi Tenant Cloud Native Environment