#19 Netflix

Here’s a complete, time-boxed, 1-hour interview-ready answer for designing a Netflix-like video streaming system. It follows your system design interview structure, including functional & non-functional requirements, APIs/data model, architecture, deep dive, and trade-offs.

---

0 – 5 min — Problem recap, scope & assumptions

Goal: Design a scalable, high-availability video streaming platform (like Netflix) that allows users to stream movies/TV shows, supports content recommendations, handles millions of concurrent users, and provides high-quality video playback.

Scope for interview:

• User registration, login, and profile management.

• Video catalog (movies, shows, episodes).

• Streaming video content on-demand.

• Recommendations and personalized content.

• High scalability and fault-tolerance.

• Analytics for content popularity, watch history, QoS.

Assumptions:

• Peak concurrent users: 10M.

• Average video size: 1–2 GB.

• Users can stream HD/4K content.

• Multi-device support (TV, mobile, web).

• Content stored in CDN-backed object storage.

---

5 – 15 min — Functional & Non-Functional Requirements

Functional Requirements

Must

1. User management: Register, login, manage profiles.

2. Video catalog: Browse/search movies, shows, episodes.

3. Video streaming: Play video on-demand, pause, resume.

4. Content delivery: Support adaptive bitrate streaming (ABR).

5. Watch history: Track progress, resume playback.

6. Recommendations: Personalized suggestions based on user behavior.

7. Rating & reviews: Users can rate content.

8. Subscriptions & billing: Access control based on subscription plan.

Should

• Multi-language support (audio/subtitles).

• Offline downloads.

• Multiple profiles per account.

Nice-to-have

• Social features (share content, friends’ watch lists).

• Live streaming for events.

• Advanced analytics for content creators.

---

Non-Functional Requirements

• Availability: 99.99% uptime, global reach.

• Scalability: Handle millions of concurrent users globally.

• Latency: Minimal startup delay (<2 sec).

• Throughput: High data throughput for video streaming (CDN + origin).

• Durability: Reliable video storage; backups.

• Consistency: Eventual consistency acceptable for recommendations; strong consistency for subscriptions/billing.

• Security: Authentication, DRM, prevent piracy.

• Monitoring & observability: Track errors, buffering, QoS, recommendations, engagement.

---

15 – 25 min — API / Data Model

APIs

User Service

register_user(username, password, email)
login(username, password) -> token
get_user_profile(user_id)
update_user_profile(user_id, profile_data)

Video Service

get_video(video_id) -> video_metadata
search_video(query, filters) -> list[video_metadata]
stream_video(video_id, quality, user_id) -> video_stream_url
get_watch_history(user_id) -> list[video_progress]

Recommendation Service

get_recommendations(user_id, count=10) -> list[video_metadata]
update_user_activity(user_id, video_id, watch_time)

Billing Service

subscribe(user_id, plan)
get_subscription_status(user_id)

Data Models

User

{
  "user_id": "string",
  "name": "string",
  "email": "string",
  "password_hash": "string",
  "profiles": [{"profile_id": "string", "name": "string"}],
  "subscription": "plan_id"
}

Video

{
  "video_id": "string",
  "title": "string",
  "description": "string",
  "duration": int,
  "genres": ["Action", "Comedy"],
  "languages": ["EN", "ES"],
  "rating": float,
  "url": "cdn_url",
  "thumbnails": ["url1","url2"]
}

Watch History

{
  "user_id": "string",
  "video_id": "string",
  "last_watched_position": int,
  "last_watched_ts": "timestamp"
}

---

25 – 40 min — High-level architecture & data flow

+------------------+       +------------------+       +------------------+
| User Client      | <---> | API Gateway      | <---> | User Service     |
| (Web/Mobile/TV)  |       +------------------+       +------------------+
|                  |                               
|                  |       +------------------+       +------------------+
|                  | <---> | Video Service    | <---> | Video Storage    |
|                  |       +------------------+       | (Object Store + |
|                  |                                | CDN)              |
|                  |       +------------------+       +------------------+
|                  | <---> | Recommendation   | <---> | Analytics DB     |
|                  |       +------------------+       +------------------+

Components

1. API Gateway: Single entry for clients; routes requests to microservices.

2. User Service: Handles authentication, profile, subscription management.

3. Video Service: Streams videos using pre-signed URLs or CDN-backed streaming.

4. Recommendation Service: ML-based or heuristic recommendations; updates based on watch history.

5. Video Storage & CDN: Store original videos and distribute via global CDN for low latency.

6. Analytics Service: Collects viewing metrics, engagement, errors, QoS.

7. Billing Service: Manages subscriptions, payments, access control.

Data Flow

1. User requests video → API Gateway → Video Service → CDN URL returned.

2. Client starts streaming; adaptive bitrate adjusts quality.

3. Video progress sent to Watch History Service.

4. Recommendation Service updates user preferences and serves suggestions.

---

40 – 50 min — Deep dive — scalability, caching, streaming

Video streaming & CDN

• Videos stored in object storage (S3, GCS) and served via CDN.

• Adaptive Bitrate Streaming (HLS/DASH) for dynamic quality adjustment.

• Use pre-signed URLs or token-based access for secure streaming.

Recommendations

• Collaborative filtering or content-based recommendations.

• Batch updates or real-time streaming using Kafka for events.

• Caching hot recommendations for fast retrieval.

Caching & performance

• Cache metadata, thumbnails, popular videos in Redis/Edge caches.

• CDN ensures low-latency global streaming.

• Load balancing across Video Service instances.

Scaling & reliability

• Microservices scaled horizontally using Kubernetes or auto-scaling groups.

• Shard user data by user_id; shard videos by video_id.

• Replicate video storage across regions for disaster recovery.

---

50 – 55 min — Back-of-the-envelope calculations

Assumptions

• Peak concurrent users: 10M, average 2 Mbps per stream → 20 Tbps peak load.

• 1M new videos/year, average size 2 GB → 2 PB/year storage requirement.

• Cache top 1% popular videos in CDN → reduces load on origin.

Storage

• Object storage: S3/Blob Storage.

• Database: SQL or NoSQL for metadata (user, watch history, subscriptions).

• Analytics: Data warehouse for ML & reporting.

Throughput

• CDN handles majority of streaming load; origin only serves cache misses or new uploads.

---

55 – 58 min — Monitoring & ops

Monitoring

• Streaming QoS: buffering, latency, playback errors.

• API metrics: request latency, errors, user activity.

• System health: service uptime, storage capacity, CDN metrics.

Operational concerns

• Handle region failovers.

• DRM enforcement for content.

• Automatic transcoding for different resolutions/bitrates.

---

58 – 60 min — Trade-offs, evolution & summary

Trade-offs

• Centralized vs distributed video storage: Distributed needed for global scale and low latency.

• Real-time vs batch recommendation: Real-time more engaging but resource-intensive.

• SQL vs NoSQL: SQL for relational data (subscriptions, profiles); NoSQL for high throughput events, watch history.

Evolution

1. MVP: Basic video upload, streaming, metadata, simple recommendations.

2. Phase 2: Adaptive bitrate, CDN, analytics, caching hot content.

3. Phase 3: ML-based personalized recommendations, multi-device sync, offline downloads.

Summary

Design a scalable, high-performance video streaming system:

• Store videos in object storage + CDN for global access.

• Microservices handle user management, video streaming, recommendations, and billing.

• Real-time updates for watch history; caching and CDN for low-latency playback.

• Scalable architecture supports millions of concurrent users and PB-scale storage.

---

If you want, I can next create a sequence diagram showing a user streaming a video, updating watch history, and receiving recommendations, which is very handy to explain in an interview.

Do you want me to create that diagram next?