CDN

Overview

A Content Delivery Network (CDN) is a geographically distributed system of cache servers that delivers web content to users with high availability and low latency. CDNs reduce load on origin servers, decrease latency by serving content from edge locations close to users, and improve reliability through replication and routing.

Why use a CDN?

• Reduced latency: Content is served from edge servers geographically closer to users
• Lower origin load: Static assets are offloaded from your origin servers
• High availability: Distributed architecture provides redundancy and fault tolerance
• DDoS protection: CDNs can absorb and mitigate large-scale attacks
• Bandwidth savings: Caching reduces repeated data transfer from origin

How CDNs work

User Request → DNS Resolution → Nearest Edge Server → Cache Hit/Miss → Response
                                        ↓
                                   Cache Miss
                                        ↓
                                 Origin Server

1. DNS routing: User's DNS request is resolved to the nearest edge server using GeoDNS or Anycast
2. Edge caching: CDN caches static assets at edge servers in multiple regions
3. Cache miss flow: On a cache miss, the edge fetches from origin, caches it, then returns to user
4. Request routing: CDNs use DNS-based routing, Anycast, or application-layer routing

Types of content served

CDN architectures

Push CDN

Content is explicitly uploaded to the CDN before users request it.

Best for: Video platforms, software distribution, large media files

Origin Server ---(upload)---> CDN Storage ---> Edge Servers ---> Users

Pull CDN (Origin-Pull)

CDN fetches content from origin on first request and caches it.

Best for: Websites, APIs, frequently changing content

User ---> Edge Server ---(cache miss)---> Origin Server
                ↓
          (cache & serve)
                ↓
            User

Multi-CDN

Using multiple CDN providers for redundancy, performance, or cost optimization.

         ┌─── CDN A (Primary)
Traffic ─┼─── CDN B (Failover)
         └─── CDN C (Regional)

Caching strategies

Cache-Control headers

Cache-Control: public, max-age=31536000, immutable
Cache-Control: private, no-cache
Cache-Control: public, max-age=3600, stale-while-revalidate=86400

Key caching concepts

• TTL (Time-to-Live): How long content stays cached before revalidation
• Cache invalidation/Purge: Force removal of cached content when origin changes
• Stale-while-revalidate: Serve stale content while fetching fresh copy in background
• Stale-if-error: Serve stale content if origin is unavailable
• Cache hierarchy: Edge → Regional → Origin (reduces origin load)

Cache key design

Cache Key = URL + Vary Headers + Query Params (selective)

Example:
/api/products?category=shoes
Vary: Accept-Language, Accept-Encoding

Performance optimizations

• Edge proximity: Serve from servers closest to users (reduces RTT)
• Connection reuse: Keep-alive connections, HTTP/2 multiplexing
• Protocol optimizations: HTTP/2, HTTP/3 (QUIC), TLS 1.3
• Compression: Gzip, Brotli at the edge
• Image optimization: WebP/AVIF conversion, resizing at edge
• Prefetching: Predictive content loading

Consistency considerations

CDNs trade immediate consistency for availability and performance.

Best practice: Use versioned/fingerprinted URLs for assets:

/static/app.a1b2c3d4.js  ← Cache forever, new version = new URL

Security features

• DDoS mitigation: Absorb volumetric attacks across distributed edge
• TLS termination: Terminate TLS at edge, reduce origin CPU
• WAF (Web Application Firewall): Filter malicious requests at edge
• Bot management: Detect and block automated threats
• Origin shielding: Hide origin IP, reduce direct attacks
• Token authentication: Signed URLs for protected content

Operational considerations

Key metrics to monitor

• Cache hit ratio: Target > 90% for static assets
• Origin traffic: Should be minimal with good caching
• Latency (p50, p95, p99): Edge latency should be < 50ms
• Error rates: 4xx/5xx from edge and origin
• Bandwidth: Egress costs and capacity

Cost optimization

• Maximize cache hit ratio (reduces origin egress)
• Use appropriate TTLs (longer = cheaper)
• Compress assets (reduces bandwidth)
• Consider reserved capacity for predictable traffic

Example: CDN for a web application

┌─────────────────────────────────────────────────────────────┐
│                        Internet                              │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                     CDN Edge Layer                           │
│  ┌─────────┐  ┌─────────┐  ┌─────────┐  ┌─────────┐        │
│  │ Edge US │  │ Edge EU │  │Edge Asia│  │Edge SA  │        │
│  └─────────┘  └─────────┘  └─────────┘  └─────────┘        │
│       Static assets, API cache, WAF, DDoS protection        │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼ (cache miss only)
┌─────────────────────────────────────────────────────────────┐
│                     Origin Infrastructure                    │
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────┐  │
│  │  Web Servers │    │  API Servers │    │ Object Store │  │
│  └──────────────┘    └──────────────┘    └──────────────┘  │
└─────────────────────────────────────────────────────────────┘

Popular CDN providers

Further reading

• High Performance Browser Networking — Ilya Grigorik
• CDN provider documentation (Cloudflare, Fastly, AWS CloudFront)
• HTTP Caching (MDN Web Docs)