Reconstructing Ctrip's Payment Engine: Architecture, Services, and Operational Practices

Hong Guangming, Development Director of Ctrip's Finance Division, introduces his work on rebuilding the company's payment backend system.

The existing payment platform, after years of growth, struggles to handle daily millions of transactions and annual trillion‑level volume, suffering from a lack of unified payment flow design, missing independent accounting, heavy reliance on stored procedures, inflexible database design, and complex mesh‑like service calls.

The redesign adopts a core concept of separating the system into two main services: a Payment Plan Service responsible for request validation, acquisition, routing, risk checks, and plan generation; and a Payment Engine Service handling plan execution, channel integration, and result notification.

Each payment channel (e.g., bank cards, third‑party providers, gift cards, wallet balance) is encapsulated in its own dedicated service.

The new overall architecture is illustrated in the following diagram.

A typical WeChat/Alipay payment flow under the new system is shown, demonstrating a simpler and clearer inter‑service call structure compared to the legacy system.

The presentation also introduces the settlement reconciliation system, which matches multi‑channel transaction and settlement data against internal accounting records, quickly identifies discrepancies, and processes nearly a hundred channel reconciliations daily, greatly enhancing operational efficiency and fund safety.

Given the high stability requirements of payment services, weekly availability metrics are presented, followed by three layers of improvement measures:

Architectural: design for high availability from the start, separate order and payment, allow multiple payment attempts per order, convert synchronous payments to asynchronous for automatic retries.

Technical: adopt distributed services, high‑availability architectures, eliminate single points of failure, improve fault tolerance.

Operational: build comprehensive monitoring and alerting, provide accurate predictive baselines for critical services, enable rapid detection and response to anomalies.

The talk concludes with a summary of the Ctrip payment engine reconstruction project.