CDN — Content Delivery Network

Overview

A CDN is a geographically distributed network of servers (edge nodes) that delivers content to users from the nearest location. CDNs reduce latency, improve availability, and offload origin servers.

Why CDNs Matter

Latency: Users get content from a nearby edge server (milliseconds vs hundreds of milliseconds)
Availability: If one edge server fails, another serves the content
Bandwidth: Origin server handles less traffic (edge servers cache)
DDoS protection: Distributed infrastructure absorbs attacks
Global reach: Content served from 100+ locations worldwide

How a CDN Works

sequenceDiagram
    participant U as User (Tokyo)
    participant DNS as DNS Resolver
    participant CDN as CDN Edge (Tokyo)
    participant O as Origin (US)
    U->>DNS: Resolve cdn.example.com
    DNS->>DNS: CDN returns nearest edge IP
    DNS->>U: IP of Tokyo edge server
    U->>CDN: HTTP GET /image.jpg
    alt Cache Hit
        CDN->>U: Return cached content
    else Cache Miss
        CDN->>O: Fetch from origin
        O->>CDN: Return content
        CDN->>CDN: Cache content
        CDN->>U: Return content
    end

CDN Architecture

graph TD
    subgraph "Edge Layer (100+ PoPs)"
        E1[Edge Server US]
        E2[Edge Server EU]
        E3[Edge Server Asia]
    end
    subgraph "Origin Shield"
        SH[Shield / Mid-Tier Cache]
    end
    subgraph "Origin"
        O[Origin Server]
    end
    E1 -->|Miss| SH
    E2 -->|Miss| SH
    E3 -->|Miss| SH
    SH -->|Miss| O

Origin Shield: An intermediate cache between edge and origin. Reduces origin load by consolidating requests from multiple edge servers.

CDN Caching

Cache-Control Headers

Cache-Control: public, max-age=3600

Directive	Meaning
public	Can be cached by CDN and browser
private	Browser-only cache (not CDN)
no-cache	Must revalidate with origin before using cache
no-store	Don’t cache at all
max-age=3600	Cache for 3600 seconds
s-maxage=3600	CDN-specific max age
immutable	Content never changes (use with versioned URLs)

Cache Key

The CDN uses a cache key to identify content:

Cache key = URL + Query string + Host header + Vary headers

Example:

https://cdn.example.com/image.jpg?v=1 → Cache entry 1
https://cdn.example.com/image.jpg?v=2 → Cache entry 2 (different)

Cache Invalidation

Method	Description
TTL expiration	Content expires after max-age
Purge	Explicitly remove from cache (API call)
Versioned URLs	`/style.css?v=2` → new URL = new cache entry
Stale-while-revalidate	Serve stale content while fetching fresh

CDN Use Cases

graph TD
    A[CDN Use Cases] --> B[Static Content]
    A --> C[Video Streaming]
    A --> D[Software Downloads]
    A --> E[API Acceleration]
    A --> F[DDoS Protection]
    A --> G[Web Application Firewall]
    B --> H[Images, CSS, JS]
    C --> I[HLS/DASH segments]
    D --> J[OS updates, game patches]
    E --> K[API responses at edge]
    F --> L[Absorb volumetric attacks]
    G --> M[WAF at edge]

DNS-Based CDN Routing

sequenceDiagram
    participant U as User
    participant LDNS as Local DNS
    participant CDN_DNS as CDN DNS
    participant E1 as Edge US
    participant E2 as Edge EU
    U->>LDNS: Resolve cdn.example.com
    LDNS->>CDN_DNS: Query (anycast)
    CDN_DNS->>CDN_DNS: Determine nearest edge
    CDN_DNS->>LDNS: Return IP of Edge EU
    LDNS->>U: IP of Edge EU
    U->>E2: Connect to Edge EU

Techniques:

Anycast: Same IP advertised from multiple locations; BGP routes to nearest
GeoDNS: DNS returns IP based on client’s geographic location
Latency-based: DNS measures latency to each edge and returns the closest

CDN Providers

Provider	Notable Features
Cloudflare	Anycast, WAF, DDoS protection, Workers (edge compute)
AWS CloudFront	Deep AWS integration, Lambda@Edge
Akamai	Largest network, enterprise-focused
Fastly	VCL configuration, real-time purging
Google Cloud CDN	GCP integration, premium tier

Interview Questions

Q: How does a CDN reduce latency? A: By caching content at edge servers geographically close to users. Instead of fetching from an origin server (possibly thousands of miles away), users get content from a nearby edge server in milliseconds.
Q: What is cache hit ratio and why does it matter? A: The percentage of requests served from cache vs origin. Higher = better performance and lower origin load. Typical targets: 80-95%. Measured by monitoring cache HIT/MISS headers.
Q: What is origin shielding? A: An intermediate cache layer between edge servers and the origin. When multiple edge servers miss the cache, they request from the shield (not directly from origin), reducing origin load.
Q: How do you invalidate CDN cache? A: Three methods: 1) Wait for TTL expiration, 2) API purge (immediate but has propagation delay), 3) Versioned URLs (change the URL for new content). Versioned URLs are the most reliable.
Q: What’s the difference between CDN and load balancer? A: A CDN is geographically distributed and caches content at the edge. A load balancer distributes traffic within a single data center. CDNs use load balancers internally, but they serve different purposes.
Q: Can a CDN cache dynamic content? A: Traditional CDNs cache static content. Modern CDNs can cache API responses (edge computing), personalized content (ESI — Edge Side Includes), and use edge compute (Cloudflare Workers, Lambda@Edge) for dynamic processing at the edge.

Common Mistakes

Not setting proper Cache-Control headers (content never cached or cached too long)
Cache key too broad (different users get same cached response) or too narrow (low hit ratio)
Not using versioned URLs for cache invalidation
Forgetting to purge CDN cache during deployments
Assuming CDN makes origin server unnecessary (origin is still needed for cache misses)

CDNs distribute content globally via edge servers, reducing latency and improving availability. Key concepts: cache hit ratio, Cache-Control headers, cache invalidation, origin shielding, and DNS-based routing. CDNs are essential for any application with a global user base.

Placement Preparation Knowledge Base