Intermediate

CDN & Edge Caching

Caching Patterns

Serve static and cacheable dynamic content from edge POPs for lower latency

Core Idea

Use globally distributed edge caches and HTTP caching semantics to deliver content close to users. Combine with edge compute for rewrites, image optimization, and security enforcement.

When to Use

When you have global users, large static content, or want to offload origin and reduce egress/latency.

Recognition Cues

Indicators that this pattern might be the right solution

High TTFB for far-away users
Origin bandwidth/egress costs are high
Traffic spikes overload origin
Static assets or cacheable dynamic pages

Pattern Variants & Approaches

Overview

Clients fetch from edge POPs; cache hits serve immediately, misses fetch from origin which queries backing stores.

Overview Architecture

When to Use This Variant

Global users and static assets
High egress costs at origin
Need WAF/DoS protection at edge

Use Case

Static websites, media delivery, cacheable dynamic pages with personalization controls.

Advantages

Lower TTFB and origin load
Tiered caching and shielding
Edge security controls

Implementation Example

# Example headers
Cache-Control: public, max-age=600
ETag: "content-hash"
Vary: Accept-Encoding, Authorization

Tradeoffs

Pros

Lower latency and higher throughput
Origin offload and egress savings
DDoS protection and WAF closer to users
Edge compute for dynamic personalization

Cons

Potential staleness and invalidation complexity
Debugging multi-layer caches is harder
Vendor lock-in for proprietary features
Extra layer adds failure modes

Common Pitfalls

Incorrect Cache-Control/ETag usage
Caching personalized or private data by accident
Missing Vary headers (auth, language, device)
Slow or missing purge/invalidation pipeline
Stampede on cache purge or miss

Design Considerations

Explicit caching headers, ETag/Last-Modified
Origin shield and tiered caching
Signed URLs/cookies and tokenized access
Image/video transforms and compression at edge
Purge APIs, soft TTL, and pre-warming

Real-World Examples

Cloudflare

Global anycast CDN with edge workers

300+ cities; massive throughput

Akamai

Enterprise CDN for media and software

Exabyte-scale monthly delivery

Netflix

Open Connect peering caches closer to ISPs

World's largest video CDN

Complexity Analysis

Scalability

Global POPs - Anycast network

Implementation Complexity

Medium - Multi-layer cache semantics

Cost

Low to Medium - Offsets origin costs

Caching Strategies API Gateway Rate Limiting Edge Compute