Technical Evolution and Architecture of China's 12306 Railway Ticketing System

We look at how Chinese netizens evaluate the national railway ticketing system 12306 and explore its technical background.

Initially, the ticketing process was chaotic, leading to widespread scalping. Early solutions involved reserving seats for intermediate stations without a unified network.

In the 1990s, the Harbin Railway Bureau developed the first nationwide ticket allocation system on SCO Unix using C, introducing a “抢票” (抢票) feature that let stations claim tickets before selling them to customers.

This system, built on minicomputers, used batch communication: stations downloaded ticket data to local servers, issued tickets locally, and returned unsold tickets after a set period.

Around 2000, with cheaper PCs, the system migrated to a client‑server (CS) model: the client was written in PowerBuilder or VB, the database used Sybase, and middleware was Tuxedo. Each railway bureau deployed its own system and synchronized data with a central ticketing center via ADSL‑style dial‑up links.

The architecture was a three‑layer distributed system (national railway ministry, regional bureaus, local stations) similar to banking systems, eventually consolidating into the centralized 12306 platform.

During China’s Spring Festival travel rush, an estimated 800 million ticket requests are processed, and the system aims for zero errors.

Key strengths of 12306 include extremely complex SKU management, nationwide channel interference, and massive traffic volume, requiring robust server scaling and real‑time processing.

Only under China’s unique administrative and population‑registry conditions could such a fully integrated, nationwide ticketing system be built.

The article also notes that the system’s success is due to the expertise of the railway and communications ministries rather than commercial tech firms.