How Tmall Scaled Its Product Pages for Double‑11: CDN‑Driven Staticization Blueprint

In the Double‑11 shopping events, Tmall’s browsing systems (product detail pages, shop pages, etc.) must handle traffic surges many times higher than everyday loads, requiring scalable capacity, stable response times, high availability, and protection against attacks and failures.

Evolution

Since 2011, the Tmall product‑detail system has undergone three major phases: staticization, unified web caching, and full CDN staticization.

Phase 1: System Staticization

The early architecture relied on thin front‑end applications and page‑level caching, which could not meet the explosive traffic of Double‑11. To overcome CPU bottlenecks, limited horizontal scaling, and high expansion costs, a staticization strategy was adopted:

Separate static and dynamic content, caching static HTML.

Introduce a distributed cache (Tair) and multi‑level caching (CDN → application → browser).

Implement active and automatic cache invalidation mechanisms.

Fill dynamic parts (price, inventory, promotions) via asynchronous CSI calls.

Phase 2: Unified Web Cache

After the first phase proved effective, a unified caching layer was built to share static resources across multiple browsing services, reducing cost and improving hit rates. Key technical components include:

Unified cache entry point using a custom Nginx‑based server (Tengine).

Support for batch and single‑object cache invalidation via a centralized purge center.

Consistent hashing for request distribution and high‑speed 10 Gbps network links to eliminate bottlenecks.

Phase 3: CDN Staticization

The unified cache removed single‑machine memory limits, enabling the final migration of static pages to a CDN. This added three technical challenges:

Distributed CDN node invalidation – solved with active invalidation requests propagated to all nodes.

Maintaining high cache hit rates – achieved by controlled node deployment and consistent‑hash rules.

Dynamic content timing – handled by ESI includes and asynchronous API calls for price, inventory, and timed banner updates.

The resulting CDN‑based static architecture dramatically increased QPS (over 7× improvement in 2012 Double‑11), reduced resource usage to half, and provided robust protection against spikes and attacks, establishing a scalable, high‑availability solution for future high‑traffic e‑commerce events.