Design a Content Delivery for Vercel

Question

Design a low-latency content delivery system that handles millions of requests. Discuss trade-offs in consistency, availability, and performance.

Codemia · Accepted Answer

### Functional Requirements
1. **Low-Latency Content Delivery**: The system must deliver content (HTML, CSS, JS, images) with latency under 100ms for 95% of requests.
2. **Dynamic Content Handling**: Support for serverless functions to return dynamic content based on user requests.
3. **Cache Invalidation**: Mechanism to manage cache invalidation for frequently updated content.
4. **Global Distribution**: Content must be served from multiple regions to provide redundancy and reduce latency.
5. **User Authentication**: Secure access to private content, supporting OAuth and JWT tokens.

### Non-Functional Requirements
1. **Scalability**: The system should handle millions of concurrent requests and scale automatically based on traffic patterns.
2. **Availability**: Target 99.99% uptime, ensuring the system is resilient to failures.
3. **Monitoring and Logging**: Implement comprehensive logging and monitoring to track performance and diagnose issues in real-time.
4. **Security**: Ensure content delivery is secured using HTTPS and protect against DDoS attacks.

Assuming Vercel has around **10 million active users** and each user makes approximately **100 requests per month**:
- **Total Requests**: 10M users * 100 requests/user = **1 billion requests/month**.
- **Daily Requests**: 1 billion / 30 days = **33.3 million requests/day** or approximately **400 requests/second** on average.
- **Peak Load**: Assuming a 10x peak load multiplier, we need to handle **4,000 requests/second** at peak.

### Storage Requirements
- If we assume an average of **1 KB** per request (including metadata), we would need:
  - **4,000 requests/sec * 1 KB/request * 60 sec/min * 60 min/hr * 24 hr/day = ~ 345 GB/day**.
- Over a month, this results in approximately **10.35 TB/month** of data, including logs and content.

### Component Diagram
- **Edge Network**: Utilize a CDN like Cloudflare to cache static assets globally.
- **Load Balancer**: Route requests to different regional services based on latency (AWS ELB).
- **Application Servers**: Serverless functions (AWS Lambda) to handle dynamic requests and API endpoints.
- **Cache Layer**: Redis for caching frequently accessed dynamic content and session data.
- **Database**: Use a multi-region database like Amazon Aurora with read replicas to ensure low-latency access and high availability.
- **Monitoring Tools**: Integrate tools like Prometheus and Grafana for monitoring and visualization of system performance.

### Technology Choices
- **CDN**: Cloudflare
- **Load Balancer**: AWS Elastic Load Balancing
- **Serverless Functions**: AWS Lambda
- **Caching**: Redis
- **Database**: Amazon Aurora (PostgreSQL)

### Database Schema Design
1. **Users Table**:
   - `user_id (UUID) PRIMARY KEY`
   - `email (VARCHAR UNIQUE)`
   - `created_at (TIMESTAMP)`
   - `updated_at (TIMESTAMP)`

2. **Content Table**:
   - `content_id (UUID) PRIMARY KEY`
   - `user_id (UUID, FOREIGN KEY)`
   - `content_type (ENUM)`
   - `content_data (BLOB)`
   - `cache_expiry (TIMESTAMP)`
   - `created_at (TIMESTAMP)`
   - `updated_at (TIMESTAMP)`

3. **Cache Invalidation Table**:
   - `cache_id (UUID) PRIMARY KEY`
   - `content_id (UUID, FOREIGN KEY)`
   - `invalidated_at (TIMESTAMP)`

### Access Patterns
- Most reads will be from the Content Table, with infrequent writes and updates, best suited for a read-optimized database.

### Architectural Decisions
1. **Consistency vs. Availability**: Adopting a multi-region database can lead to eventual consistency. We must balance between strong consistency for critical user data and eventual consistency for cache data to enhance availability and performance.
2. **Cache Invalidation**: We need to decide between time-based invalidation (TTL) and event-based invalidation. TTL is simpler but might serve stale content.
3. **Deployment Strategy**: We could implement a blue-green deployment strategy for serverless functions to minimize downtime and risks during updates.
4. **Monitoring and Alerting**: Trade-off between detailed metrics and operational overhead. Focus on key metrics (latency, error rates) and use alerting thresholds to minimize alert fatigue.

### Operational Excellence
- Implement chaos engineering practices to test resilience.
- Set SLOs/SLIs for latency and availability, ensuring performance metrics align with user expectations.

Design a Content Delivery for Vercel

Vercel

What the Interviewer Expects

Key Topics to Cover

How to Approach This

Possible Follow-up Questions

Practice a Similar Problem on Codemia

Sample Answer

Requirements

Functional Requirements

Capacity Estimation

Submit Your Answer

Vercel Software Engineer Interview Guide

Related Questions

Design a high-throughput Inventory Management System

Design a low-latency Rate Limiting System

Design a fault-tolerant Payment System

Design a fault-tolerant Messaging System

Design Walmart Product Search