Ctrip Hotel BFF Architecture Migration and Efficiency Boost with NestJS and Cloud Functions

The article begins by describing the challenges faced by Ctrip Hotel's front‑end BFF layer during architecture migration and efficiency improvement, emphasizing the need to separate UI concerns from domain services as user demands become more diverse.

Two BFF patterns are compared: "one‑code‑one‑end", which provides a dedicated BFF per client but leads to monolithic, duplicated code, and "one‑code‑many‑ends", which shares a single BFF across multiple clients while handling UI differences internally, resulting in higher development efficiency.

Adopting the "one‑code‑many‑ends" approach, the team selected NestJS as the foundational framework, highlighting its modularity, built‑in dependency injection, TypeScript support, decorator‑based routing, microservice capabilities, testing tools, and seamless integration with other libraries.

To address the complexities of multi‑client BFF maintenance, a multi‑end strategy pattern was introduced, separating per‑client logic into independent strategy instances, reducing code coupling, enabling parallel execution, and simplifying view‑model consistency.

The article then details Ctrip's Node.js cloud‑function platform, covering its lightweight runtime, middleware ecosystem, trigger mechanisms, layer management, elastic scaling, blue‑green and gray‑release deployment, and monitoring tools, all of which accelerate function development and improve resource utilization.

By packaging the NestJS‑based BFF as a cloud‑function template, the team achieved faster cold‑start times, lower memory consumption, and better performance under identical QPS conditions.

Finally, a dynamic gateway layer is introduced to handle inter‑API composition, whitelist, gray‑release, and scenario‑based customization, further decoupling multi‑end view logic and enabling cross‑application UI dynamism.