How We Boosted CDN Stability by 80% with Automated Testing and Dual Architecture

Introduction

As test development engineers, we ensure online service quality; CDN stability directly impacts user experience. Recent challenges in CDN testing were overcome with automation, and the international CDN architecture offers new insights.

International CDN Dual Architecture

Full Overseas Architecture

All resources are stored in S3 in Singapore and Oregon, delivered via AWS CDN → AWS S3. Korea uses AWS CDN → Tencent Cloud overseas CDN → Tencent Cloud COS (Shanghai) to balance performance and cost.

Dynamic Overseas Architecture

For large amounts of resources stored in Tencent Cloud that are hard to migrate, we asynchronously move data to overseas nodes on first access, using AWS CDN → Tencent Cloud overseas CDN → Tencent Cloud COS to ensure efficient distribution.

Why CDN Stability Is Critical for Test Development

High‑frequency change risk : frequent online config changes make manual regression inefficient.

Data black box : lack of systematic CDN image detection leads to delayed issue discovery.

User experience impact : image loading failures cause page stalls and affect key business metrics.

These challenges require automated testing and data monitoring.

Automated Backend Interface Validation

Core Idea

Focus on representative “open scheme” scenarios covering most CDN domains.

Filter requests by resourceType to capture image requests.

Validate HTTP status codes (200/206/304) and flag anomalies.

Core Code Implementation

Define domain prefixes and arrays for allUrls and non200Urls.

Listen to response events, collect URLs and status codes.

Log counts and assert that at least one URL is matched and all have acceptable status codes.

Front‑End Image Detection: Data‑Driven Quality Guard

We built a pipeline for data collection, rule validation, and automatic alerts.

Technical Solution

Automated data collection using evaluate to get natural size, element size, etc.

Intelligent analysis model calculates aspect ratio, file size, and detects abnormal cropping or oversized images.

Real‑time alert via robot pushes anomalies to developers.

Core Functions

a. getImgAndInput – finds

and background‑image elements, filters by domain (kujiale, coohom), fetches image info, and records URLs.

b. getImageInfo – uses img.evaluate to obtain src, natural dimensions, element dimensions, and sends a HEAD request to get content‑length and content‑type.