edge-computing

name	edge-computing
description	Cloudflare Workers, Deno Deploy, Vercel Edge Functions, edge patterns (geo-routing, caching). Use when implementing edge compute, CDN logic, or global low-latency APIs.

Edge Computing — Cloudflare Workers, Deno Deploy

Compute global faible latence, proche utilisateur.

Platforms

Cloudflare Workers — V8, 300+ POPs, KV/Durable
Deno Deploy — Deno, 30+ POPs, TypeScript
Vercel Edge — V8, 20+ POPs, Next.js
Fastly Compute@Edge — WASM, 70+ POPs

Geo-Routing

export default {
  async fetch(req) {
    const api = req.cf.country === 'US' ? 'us-api' : 'eu-api';
    return fetch(`https://${api}.example.com${req.url}`);
  }
};

Edge Caching (KV)

const key = new URL(req.url).pathname;
let cached = await env.KV.get(key);
if (cached) return new Response(cached, { headers: { 'X-Cache': 'HIT' } });
const resp = await fetch('https://origin.com' + key);
const body = await resp.text();
await env.KV.put(key, body, { expirationTtl: 3600 });
return new Response(body, { headers: { 'X-Cache': 'MISS' } });

A/B Testing

const variant = Math.random() < 0.5 ? 'A' : 'B';
resp.cookies.set('ab_test', variant);

Bot Detection

if (/bot|crawler/i.test(req.headers.get('User-Agent'))) {
  return new Response('Forbidden', { status: 403 });
}

Durable Objects (Stateful)

export class ChatRoom {
  async fetch(req) {
    const [client, server] = Object.values(new WebSocketPair());
    this.sessions.push(server); server.accept();
    server.on('message', e => this.broadcast(e.data));
    return new Response(null, { status: 101, webSocket: client });
  }
}

Constraints

CPU 10-50ms → offload heavy
Memory 128MB → streaming
No FS → KV/R2

---

Voir @.claude/skills/wasm/SKILL.md