Edge Computing & CDN Cache Invalidation (Surrogate-Keys)

Q: What is the difference between `Cache-Control` and `Surrogate-Control` headers?

`Cache-Control` instructs end-user browser caches; `Surrogate-Control` instructs intermediate CDN reverse proxies and is stripped before the response reaches the browser.

Q: How do Edge Workers (Cloudflare Workers / Fastly VCL) complement CDN caching?

Edge Workers execute custom JavaScript/Rust logic directly on CDN edge nodes, handling A/B testing splits, geolocation redirects, and JWT authentication before requests ever reach the origin.

Q: What is the danger of purging a high-cardinality tag like 'homepage' during flash sales?

It triggers an Origin Stampede (Cache Stampede), where thousands of incoming requests bypass the CDN cache simultaneously and overwhelm the origin database.

Stack: SOFTWARE ARCHITECTURE STACK →Staff (L6-L7)pattern

⚡THE SHORT ANSWER

By tagging dynamic HTTP responses with granular metadata identifiers (Surrogate-Key: product_123 category_45) and purging specific tags via edge CDN APIs in <150ms while serving stale cached data in the background via stale-while-revalidate.

Engineering Handbook & Failure Dynamics

6-Dimensional Architecture Breakdown

⚙️1. Underlying Mechanism

Execution

Traditional CDN caching relies on exact URL paths with fixed TTLs. If a merchant updates a product price, purging by URL is impossible if the product appears on 500 different category and search pages. Surrogate-Keys (Cache-Tags) solve this: when the origin renders a page, it attaches a header listing all entities on that page (e.g. Surrogate-Key: prod_982 brand_nike cat_shoes). The CDN indexes the cached response against these keys and strips the header before serving users. When product #982 updates, the origin calls the CDN Purge API for key prod_982. Within 150ms, all 500 edge nodes globally invalidate every cached page containing that product. Combining this with stale-while-revalidate ensures users receive instant responses while the CDN refreshes content asynchronously from the origin shield.

🎯2. Appropriate Use Context

Scope

High-traffic media sites, e-commerce catalog pages, dynamic content publishers, and multi-region APIs requiring sub-50ms global TTFB (Time to First Byte) with real-time freshness.

⚠️3. Production Failure Modes

P0 Risk

✓
1. ▸Wildcard Purge Stampede: Executing a global 'Purge Everything (*)' API call during peak hours, driving 200,000 requests/sec directly to the origin database and causing an immediate outage
✓
1. ▸Stale Dynamic Price Leak: Caching user-specific cart prices on shared CDN nodes without proper Vary: Cookie or private headers
✓
1. ▸Purge Rate Limit Exhaustion: Firing 10,000 purge API calls/minute during a bulk product sync, getting blocked by CDN rate limits

📡4. Diagnostic Signals & Telemetry

Telemetry

CDN Edge Cache Hit Ratio (target >92%), Origin Shield bandwidth offload percentage, Purge API execution duration (P99 time to global edge eviction), and Origin server CPU utilization spikes.

🛡️5. Prevention & Safeguards

Safeguards

✓
Enforce Surrogate-Control headers with stale-while-revalidate and stale-if-error
✓
deploy an Origin Shield caching tier between CDN edges and backend origin
✓
batch tag purges (up to 256 keys per API call)
✓
and always include unique asset hash digests in static asset URLs

⚖️6. Architectural Trade-offs

Trade-off

Delivers near-instantaneous global response times (<20ms TTFB) and reduces origin database load by 95% at the cost of implementing tag generation discipline and managing CDN purge API dependencies.

📋

Case Study (TinyCTO In-Field Example)

REAL-WORLD TELEMETRY

TinyCTO Incident 083: A breaking news homepage had a 60-second fixed TTL. During an election night, an incorrect headline remained visible for 59 seconds despite editor corrections, sparking a media crisis. Integrating Fastly/Cloudflare Surrogate-Keys allowed the CMS to purge story_582 in 120ms globally, ensuring instant editorial corrections with a 98.4% edge cache hit ratio.

Interactive Concept Drills

3 Cards

How do Surrogate-Keys (Cache-Tags) revolutionize CDN cache invalidation?

They decouple cache invalidation from URL paths, allowing an origin to purge thousands of disparate web pages simultaneously with a single tag identifier (e.g. `product_123`).

What does the `stale-while-revalidate` HTTP cache directive do?

It instructs the CDN to immediately serve an expired (stale) cached asset to the user in 10ms, while asynchronously revalidating and refreshing the cache from the origin in the background.

What is an 'Origin Shield' in a global CDN architecture?

A centralized, designated caching tier placed between hundreds of global edge PoPs and the origin server, coalescing redundant edge requests into a single origin fetch.