Edge Computing Explained: Why Your Code Is Moving Closer to Users

Edge computing is going from buzzword to architecture reality. Here's what it actually means and why you should care.

The Problem Edge Solves

Traditional web app request flow:

User in Mumbai 
  → Request to server in Virginia (200ms)
  → Server processes request
  → Response back to Mumbai (200ms)
  → Total: 400ms+ latency

Edge computing puts compute closer:

User in Mumbai
  → Request to Mumbai edge node (~5ms)
  → Edge runs your code
  → Response (~5ms)
  → Total: 10-20ms latency

This isn't theoretical — it's 20x real-world improvement.

What "Edge" Means in Practice

Edge != CDN (though CDNs were the first edge)

Cloudflare Workers: The Leading Platform

// Runs in 300+ locations globally
// Cold start: ~0ms (always warm)
 
export default {
  async fetch(request, env, ctx) {
    const url = new URL(request.url);
    
    // A/B test at the edge — no origin needed
    if (Math.random() < 0.5) {
      return new Response('Variant A');
    }
    
    // Geo-routing at the edge
    const country = request.cf?.country;
    if (country === 'IN') {
      return Response.redirect('https://in.yourapp.com');
    }
    
    return fetch(request); // Pass through to origin
  }
};

Real Edge Use Cases

• Personalization — Show different content based on user's country/device without origin hits
• Auth middleware — Verify JWTs at the edge before requests reach your server
• A/B testing — Split traffic without JS flicker
• Rate limiting — Block bad actors before they reach your origin
• Response caching — Cache personalized responses (edge state)
• API aggregation — Combine multiple backend calls at the edge

The Limitations

• Restricted runtime — No Node.js APIs. Workers use Web APIs only.
• Limited CPU time — 50ms CPU time limit per request on free tier
• No stateful connections — WebSockets need workarounds (Durable Objects)
• Cold edge data — Edge KV stores have eventual consistency

Is It Worth the Complexity?

For global apps with latency-sensitive operations: Yes. For apps with 90%+ of users in one region: Often not.

Edge is a tool, not a default architecture. Apply it where latency actually matters.