edge-computing

name	edge-computing
description	Edge functions, serverless edge, Cloudflare Workers, and edge computing patterns for low-latency applications.
triggers	{"keywords":["edge","CDN","Cloudflare Workers","Vercel edge","middleware","edge function","serverless"]}
auto_load_when	Implementing edge functions or CDN logic
agent	devops-engineer
tools	["Read","Write","Bash"]

Edge Computing Patterns

1. When to Use Edge

Use edge when:
├── <50ms latency required
├── Static content + light processing
├── Request rate >100k/month
├── Geographic distribution needed
└── Cost reduction for simple logic

Use traditional server when:
├── Complex long-running processes
├── Large payload processing
├── WebSockets/streaming
├── Database connections heavy
└── Specialized libraries needed

2. Platform Selection

Edge platforms:
├── Cloudflare Workers: 35+ PoPs, KV, D1, Durable Objects
├── Vercel Edge: Next.js integration, Middleware
├── Netlify Edge: Functions, identity
├── Deno Deploy: Deno runtime, KV
├── AWS Lambda@Edge: CloudFront integration
└── Fastly Compute: Full compute, FQL

3. Cloudflare Workers Patterns

Architecture:
├── Workers: JavaScript/TypeScript, W3C standard
├── KV: Key-value store (eventual consistency)
├── D1: SQLite database
├── Durable Objects: State, WebSocket coordination
├── R2: S3-compatible storage, no egress fees
└── Queues: Async processing

Use cases:
├── A/B testing at edge
├── Request rewriting/redirecting
├── Auth verification (JWT)
├── API rate limiting
├── Static site enhancement
└── Bot detection

4. Cold Start Patterns

Cold start characteristics:
├── First request after idle (ms to seconds)
├── Subsequent requests: warm
├── Varies by platform

Strategies:
├── Code split: Minimal main entry
├── Keep warm: Scheduled pings
├── Prefetch: Predictive warming
└── Hybrid: Fallback to origin

Measure:
├── Track cold start latency
├── Adjust timeout accordingly
└── Accept some slow first requests

5. State Management

Stateless patterns:
├── Always possible: Embed state in request
├── URL-based: Query params, path params
├── Headers: X- prefix for custom
└── Cookies: Signed, encrypted

Stateful patterns:
├── Durable Objects (CF): Single instance
├── External DB: Connection per request
└── External cache: Redis, Memcached

Prefer stateless:
├── Better scaling
├── No affinity issues
└── Simpler operations

6. Data Patterns

Edge-optimized data:
├── Edge KV: Read-heavy, <1ms typical
├── D1: SQLite, relational needs
├── R2: Large objects, media
└── External API: Origin as source

Patterns:
├── Cache at edge: Stale-while-revalidate
├── Replicate: Multi-region
├── Read-through: Proxy to origin
└── Write-back: Async to origin

7. API Patterns

Edge API design:
├── Stateless endpoints: No server state
├── Idempotent operations: Safe retries
├── Versioned: /v1/, /v2/
├── Compressed: gzip/brotli
└── Cached: Cache-Control headers

Request handling:
├── Body parsing: Native/streaming
├── Validation: Schema validation
├── Transform: Request/response mapping
└── Route: Clean URL patterns

8. Security Patterns

Edge security:
├── Origin verification: Challenge passwords
├── Rate limiting: Token bucket, distributed
├── Bot detection: Edge-specific services
├── Auth: JWT verification edge-side
└── Secrets: Environment variables

Data protection:
├── TLS always: Edge to origin
├── Signed URLs: Private content
└── Gzip: Reduced exposure

9. Migration Patterns

Moving to edge:
├── Start with static: Cache everything possible
├── Add middleware: Edge functions for logic
├── Migrate APIs: Stateless first
├── Database last: Heavy state moves expensive

Hybrid approach:
├── Edge for public, fast
├── Server for private, complex
└── Clear split responsibility

Key Patterns

1. Stateless preferred - Embed all needed in request
2. Cache aggressively - Edge caches are powerful
3. Test latency - Measure from user locations
4. Start simple - Migrate static first
5. Handle cold starts - Accept warm-up latency
6. Watch costs - Many platforms have free tiers

---

Anti-Patterns

❌ Running Node.js-only APIs at edge (unsupported runtime)
✅ Use Edge Runtime compatible APIs: fetch, crypto, TextEncoder only

❌ Accessing database directly from edge worker
✅ Use edge-compatible DBs (PlanetScale, Neon, Turso) or cache layer

❌ Edge functions for long-running tasks (>30s limit)
✅ Edge for short-lived transformations; offload heavy work to serverless

❌ No geo-awareness — same content to every region
✅ Edge KV for regionally replicated data; vary by cf-ipcountry header

❌ Debugging edge with console.log (no runtime visibility)
✅ Use structured logging + wrangler tail / Cloudflare Logpush

---

Quick Reference

Platform	Runtime limit	Best for
Cloudflare Workers	50ms CPU	Global low-latency
Vercel Edge	25s wall clock	Next.js middleware
Deno Deploy	Generous	Full Deno APIs
Fastly Compute	50ms	CDN logic
AWS Lambda@Edge	5s	CloudFront integration

🌍 Universal Language Support

• Turkish Native: This skill natively supports Turkish. If the user prompt is in Turkish, all analysis, formatting, and output MUST be entirely in Turkish. You do not need explicit "write in Turkish" instructions.