Edge Computing

Overview

Edge computing pushes computation and data storage closer to the sources of data — at the “edge” of the network, near users. Instead of sending all data to a centralized cloud, processing happens at CDN edge nodes, IoT gateways, or regional servers.

Why Edge Computing Matters

• Latency: Processing near users = milliseconds, not hundreds of milliseconds
• Bandwidth: Only relevant data sent to cloud, reducing costs
• Privacy: Sensitive data processed locally, not transmitted
• Reliability: Works even with intermittent connectivity
• Real-time: Enables applications that need instant responses

Cloud vs Edge vs Fog

graph TD
    subgraph Cloud
        C[Cloud Data Center<br>High compute, high latency]
    end
    subgraph Fog
        F[Regional Servers<br>Medium compute, medium latency]
    end
    subgraph Edge
        E[Edge Nodes / CDN PoPs<br>Low compute, low latency]
    end
    subgraph Device
        D[IoT Devices / Phones<br>Minimal compute, zero latency]
    end
    D --> E
    E --> F
    F --> C

Edge Computing Platforms

CDN-Based Edge Compute

How Edge Compute Works

sequenceDiagram
    participant U as User
    participant E as Edge Node (Tokyo)
    participant O as Origin (US)
    U->>E: HTTP Request
    E->>E: Execute edge function
    Note over E: Process request at edge<br>(auth, routing, personalization)
    alt Data available at edge
        E->>U: Response (no origin needed)
    else Data needed from origin
        E->>O: Fetch data
        O->>E: Return data
        E->>E: Process and cache
        E->>U: Response
    end

Use Cases

graph TD
    A[Edge Use Cases] --> B[A/B Testing]
    A --> C[Authentication]
    A --> D[Request Routing]
    A --> E[Personalization]
    A --> F[Image Optimization]
    A --> G[Bot Detection]
    A --> H[API Gateway]
    A --> I[Real-time Analytics]
    B --> J[Route users to variants at edge]
    C --> K[Validate JWT without origin]
    D --> L[Route by geography/device]
    E --> M[Serve localized content]
    F --> N[Resize/compress images on-the-fly]
    G --> O[Block bots before origin]
    H --> P[Rate limit, transform APIs]
    I --> Q[Aggregate metrics at edge]

Edge Function Example (Cloudflare Workers)

export default {
  async fetch(request) {
    const url = new URL(request.url);
    
    // A/B testing at edge
    const cookie = request.headers.get('Cookie');
    const variant = cookie?.includes('variant=B') ? 'B' : 'A';
    
    // Route to different backends
    if (url.pathname.startsWith('/api/')) {
      return fetch(`https://api.example.com${url.pathname}`);
    }
    
    // Serve personalized content
    const country = request.cf?.country;
    const greeting = country === 'JP' ? 'こんにちは' : 'Hello';
    
    return new Response(`${greeting}! You're in variant ${variant}`);
  }
};

Edge vs Traditional Architecture

graph TD
    subgraph "Traditional"
        T_U[User] -->|Long distance| T_Cloud[Cloud Server]
        T_Cloud -->|Process| T_Cloud
        T_Cloud -->|Response| T_U
    end
    subgraph "Edge"
        E_U[User] -->|Short distance| E_Edge[Edge Node]
        E_Edge -->|Process locally| E_Edge
        E_Edge -->|Response| E_U
        E_Edge -.->|Async sync| E_Cloud[Cloud]
    end

Interview Questions

1. Q: What is edge computing and how does it differ from cloud computing? A: Edge computing processes data near the user (at CDN PoPs, 5G towers, IoT gateways) rather than in centralized cloud data centers. It reduces latency, saves bandwidth, and enables real-time applications. Cloud computing provides massive compute but with higher latency.

2. Q: When would you use edge computing? A: When you need: low latency (real-time gaming, AR/VR), bandwidth savings (IoT data filtering), privacy compliance (process data locally), or offline capability. Don’t use edge for heavy computation or large datasets.

3. Q: What is Cloudflare Workers? A: A serverless edge computing platform that runs JavaScript/Rust in V8 isolates at 300+ CDN edge locations. Each request runs in an isolated environment with minimal cold start (<1ms). Used for A/B testing, auth, routing, and API processing at the edge.

4. Q: What are the limitations of edge computing? A: Limited compute resources (can’t run heavy ML models), limited storage (stateless by default), cold start issues (though minimal in modern platforms), debugging complexity (distributed), and vendor lock-in.

5. Q: What’s the difference between edge and fog computing? A: Edge computing processes at the network edge (CDN, 5G tower). Fog computing is a broader concept that includes processing at any point between the device and cloud, including regional data centers. Edge is a subset of fog.

Common Mistakes

• Trying to run heavy computation at the edge (limited resources)
• Not considering data consistency (edge nodes may have stale data)
• Vendor lock-in (each platform has different APIs)
• Not understanding cold start behavior
• Over-complicating architecture (sometimes a simple cloud server is better)

Summary

Edge computing brings computation closer to users, reducing latency and bandwidth usage. CDN-based edge compute (Cloudflare Workers, Lambda@Edge) enables running code at 100+ locations globally. It’s ideal for A/B testing, authentication, routing, and personalization.

Cross-References

• CDN Overview
• How CDN Works
• Load Balancing
• 5G — Edge computing enabler
• SDN — Network programmability

Cross References

• How CDN Works
• DNS Caching
• Distributed Partitioning
• Cloud Overview