How RocketMQ Transformed Alibaba’s Logistics System for Double‑11 Peaks

Cainiao’s logistics system originated from Tmall and shared trading platforms, forming a tightly coupled relationship that has long supported the core link of the entire e‑commerce ecosystem.

With the rapid growth of online shopping and record‑breaking Double‑11 order peaks each year, technical investment and costs have risen sharply; the increasing payment capacity now matches order creation, putting additional pressure on the logistics system. The technical “umbilical cord” between transactions and logistics has become a heavy chain around Cainiao.

Analyzing 2015 Double‑11 data shows that only a few hundred thousand orders were shipped within the first hour after creation, a tiny fraction compared to payment volume, indicating that supporting massive order‑creation traffic is driven more by the need to ensure transaction‑payment‑logistics stability than by logistics demand itself.

Logistics during Double‑11 is driven by shipment documents, requiring extensive human and material resources. The physical, offline coordination has relatively stable traffic without obvious peaks, and users care mainly about successful transaction order creation, not immediate logistics order creation.

Consequently, scaling logistics infrastructure to match rising transaction order peaks each year is inefficient; investing heavily in a non‑essential scenario is unwise, making an architectural upgrade urgent.

RocketMQ – The Choice of Cainiao Architects

Facing a surge from 180,000 to 300,000 orders per second for Double‑11 2016, architect Wang Wei selected RocketMQ, Alibaba’s widely used distributed messaging middleware, to decouple the systems.

RocketMQ, battle‑tested through many Double‑11 events, offers world‑leading performance, supporting billions of messages with precise consumption rate control, making it ideal for peak‑shaving and load‑balancing.

By using RocketMQ for asynchronous decoupling, the logistics order center can maintain a stable consumption rate (e.g., 80,000 orders per second) despite a transaction peak of 500,000 orders per second, reducing labor and material costs and avoiding stability risks from synchronous dependencies. The two systems communicate efficiently while focusing on their own domains.

After two months of development, testing, and gradual rollout before Double‑11 2016, the architecture shifted from e‑commerce‑centric high concurrency to a logistics‑friendly model. Even as order‑creation peaks later grew fourfold, Cainiao’s logistics QPS remained unchanged, saving substantial technical costs and laying a foundation for future cost reductions.