How Alibaba’s Full‑Link Stress Test Powers Double 11’s Record‑Breaking Traffic

During the 2017 Double 11 shopping festival, Alibaba set new records with a peak of 325,000 transaction creations and 256,000 payments per second. The massive traffic surge was handled by a site built from thousands of business systems and technical components, thanks to the secret of “full‑link stress testing”.

Alibaba’s diverse business lines are supported by distributed technical systems, making it a major challenge to accurately assess the service capacity of the entire site, especially during large‑scale promotional events.

Since 2011, Alibaba has experimented with production‑environment single‑system stress tests to improve capacity planning. While more accurate than traditional tests, this approach is limited—like testing each car part individually without verifying the assembled vehicle, leaving the system vulnerable to failure under real load.

Full‑link stress testing solves this problem by simulating real business scenarios across the whole site, ensuring performance, capacity, and stability at every link. The Double 11 scenario is rehearsed multiple times before the actual day, so the live event is merely another run of a thoroughly validated test.

The technology rests on three core pillars:

1. Conducting full‑business pressure tests directly in the production environment.

2. Designing test scenarios that closely mirror real business workflows.

3. Building a platform capable of generating Double 11‑level peak request loads.

Ensuring that production‑environment tests do not interfere with normal traffic requires strict data isolation. Test data is stored in a dedicated “shadow” area, and all test requests carry a special marker that propagates through distributed calls, allowing middleware to route them safely without impacting live users.

Accurate business‑scenario matching is crucial for reliable results. By modeling thousands of user behavior factors—such as device mix, browsing patterns, and purchase actions—using online data and historical behavior, the intelligent data engine creates high‑fidelity test models that closely resemble real traffic.

During a full‑link test, the platform simulates billions of consumer requests. It consists of a control center (the brain) and thousands of engine nodes deployed on global CDN points, delivering up to 10 TB of test data, supporting over a hundred million concurrent users and millions of TPS.

In the 2017 preparation phase, the “Sharp Plan” upgraded the full‑link testing capability with automation and intelligence, reducing the number of required test cycles from eight to three and cutting both time and cost.

Since its inception, full‑link stress testing has become the “nuclear weapon” for Double 11 preparation, uncovering hundreds of potential system issues each year, dramatically improving site stability and reducing manual capacity‑planning effort.

In July 2017, Alibaba Cloud commercialized the internal full‑link testing platform, integrating it with the Performance Testing Service (PTS) to create the PTS Platinum edition. Customers can now perform comprehensive site performance and capacity tests and optionally engage expert services to achieve Double 11‑grade reliability.

Leveraging Alibaba’s extensive infrastructure experience, cloud‑based services such as PTS, DRDS, and EDAS bring large‑scale, high‑availability capabilities to enterprises, helping them meet rapidly changing market demands with robust, publicly available infrastructure.