System Design: Netflix‑Like Video Streaming Platform

Context

Design a large-scale video streaming platform that supports web, mobile, and TV clients for millions of concurrent viewers worldwide. The system should prioritize availability, low-latency playback, security, and cost efficiency.

Assume:

• Peak: 5–10 million concurrent streams globally.
• Target startup latency < 2 seconds; low rebuffering; 99.9%+ availability for playback.
• Mix of on-demand VOD content; live not required (optional considerations welcome).

Requirements

Design the system and cover the following components:

1. User Authentication and Profiles

• Sign-in, sessions, MFA, device limits, parental controls, multiple profiles per account.

2. Content Catalog and Metadata

• Titles, seasons/episodes, genres/tags, availability windows, rights/regions, subtitles/audio tracks.

3. Search and Personalized Recommendations

• Full-text search, browse pages, personalized rows, cold-start handling.

4. Video Ingestion and Transcoding Pipeline

• Secure uploader, QC, per-title encoding, thumbnails, captions, packaging to HLS/DASH.

5. DRM and Adaptive Bitrate Streaming

• Widevine/FairPlay/PlayReady, license services, CMAF, HLS/DASH manifests, LL-HLS optional.

6. CDN and Edge Caching Strategy

• Multi-CDN, edge authorization, origin shielding, cache warming/purges.

7. Storage and Geo-Replication

• Masters vs renditions, object storage layout, lifecycle tiers, cross-region replication.

8. Playback Session Management

• Session creation, concurrency enforcement, heartbeats, QoE metrics, resume playback.

9. APIs and Service Discovery

• External API gateway, internal gRPC, schema, rate limits, idempotency, service mesh.

10. Multi-Region Failover and Data Consistency

• Active-active vs active-passive, routing, data models with consistency choices.

11. Observability and A/B Testing

• Metrics/traces/logs, QoE SLOs, experimentation platform, guardrails.

12. Cost Optimization

• Encoding ladder strategy, codec choices, storage lifecycle, CDN offload, compute efficiency.

Deliver a high-level architecture and justify key trade-offs. Include back-of-the-envelope capacity estimates and call out pitfalls and edge cases.