CDN

Introduction

A Content Delivery Network (CDN) caches static and cacheable dynamic content at edge PoPs (points of presence) close to users, reducing latency and origin load. CloudFront, Akamai, Fastly, and Cloudflare serve images, JS/CSS, video segments, and even API responses with short TTLs.

In HLD, CDNs are mandatory at scale for read-heavy assets and global user bases. URL shortener redirects, OTT video chunks, and product catalog images all benefit. Interviewers expect you to mention cache invalidation, TTL strategy, and origin shielding.

This lesson covers CDN placement, caching headers, dynamic content acceleration, and cost considerations.

Understanding the topic

Key concepts

• Edge cache: content stored geographically; cache hit avoids round trip to origin.
• Cache key: URL path + query string rules + Vary headers (careful with auth).
• TTL: Cache-Control max-age, s-maxage for shared caches, stale-while-revalidate.
• Invalidation: purge by path or tag after deploy or content update.
• Origin shield: secondary cache layer reducing origin thundering herd.
• Dynamic site acceleration: CDN maintains persistent connections to origin for uncacheable API.

flowchart LR
  User --> Edge[CDN Edge]
  Edge -->|miss| Origin[Origin Server]

Internal architecture

Architecture overview

flowchart LR
  User --> Edge[CDN Edge]
  Edge -->|miss| Origin[Origin Server]

Step-by-step explanation

1. User DNS resolves to CDN anycast IP nearest PoP.
2. PoP cache HIT → return immediately; MISS → fetch from origin (or shield).
3. Static assets: long TTL + fingerprinted filenames (app.a1b2.js).
4. API GET with public data: short TTL (5–60s) or edge side includes.
5. Video HLS/DASH: segment files cached at edge; manifest low TTL.
6. Signed URLs/cookies for private content (premium video, paid downloads).

Informative example

CloudFront-style cache behavior and Spring Cache-Control headers for product catalog:

@RestController
@RequestMapping("/api/v1/products")
public class ProductController {
    private final ProductService products;

    public ProductController(ProductService products) {
        this.products = products;
    }

    @GetMapping("/{id}")
    public ResponseEntity<ProductDto> get(@PathVariable String id) {
        ProductDto dto = products.findById(id);
        return ResponseEntity.ok()
            .cacheControl(CacheControl.maxAge(60, TimeUnit.SECONDS).cachePublic())
            .header("Vary", "Accept-Encoding")
            .body(dto);
    }
}

// CDN behavior (conceptual YAML)
// default_ttl: 86400 for /static/*
// min_ttl: 0 for /api/* — respect origin Cache-Control

Never cache personalized or authenticated responses without Vary and private directive. URL shortener redirects cache well with long TTL.

Real-world use

Real-world use cases

• E-commerce product images and JS bundles globally.
• OTT: video segment delivery — 80%+ bandwidth from CDN.
• Social: avatar and media thumbnails at edge.
• News/fintech static marketing pages and public rate tables.

Best practices

• Fingerprint static assets for immutable long-term caching.
• Use cache tags for bulk invalidation on catalog updates.
• Monitor cache hit ratio — low ratio wastes CDN cost.
• Geo-restrict content at CDN edge for licensing compliance.
• Enable HTTP/2 and Brotli at CDN for smaller payloads.
• Protect origin with CDN-only access (secret header) preventing bypass.

Common mistakes

• Caching Set-Cookie responses — users see wrong sessions.
• Query string breaks cache key unintentionally (?v=1 on every deploy).
• No purge strategy — stale prices after promotion.
• CDN for all API traffic without TTL discipline — stale data bugs.
• Ignoring HTTPS cert management at CDN for custom domains.

Advanced interview questions

Summary

CDN caches content at edge PoPs for latency and scale. Static assets use long TTL + fingerprinting; APIs use short TTL carefully. Invalidation and cache keys prevent stale or leaked data. OTT and URL shortener designs rely heavily on CDN. Origin shield protects backend from miss storms. Data layer choices (SQL vs NoSQL) follow CDN edge caching.