Design and Implementation of a FaaS Platform for the Tongtian Tower Activity System

The Tongtian Tower platform, used for building activity pages across JD.com, JD Express, WeChat, QQ and other channels, faced challenges such as heavy reliance on third‑party data sources, fast‑changing business requirements, and high front‑back integration costs under a traditional Java monolithic architecture.

To address these issues, the team introduced a serverless Function‑as‑Service (FaaS) platform, leveraging the elasticity, rapid development, and zero‑ops characteristics of serverless computing.

The article explains the evolution of cloud service models— IaaS, CaaS, PaaS, and FaaS— and highlights FaaS as a way to let developers focus solely on business logic without managing hardware or operating systems.

The Tongtian Tower FaaS platform consists of an integrated SDK, a one‑stop operation and release console (Function Center), JD’s distributed configuration system (DUCC), and the JD File Storage System (JFS). Although the current version is not a fully elastic, centrally managed FaaS product, it serves as an initial exploration toward a true serverless solution.

Key features include low integration cost via the SDK, local function execution within the caller’s process for near‑zero latency, support for generic JSF (JD RPC) calls through annotation‑based APIs, and transparent monitoring of function deployment and runtime metrics.

The SDK design draws inspiration from MyBatis mapper concepts, mapping functions to interfaces so developers only need to define a function interface and invoke it, while the platform handles configuration and execution.

Practical use cases demonstrated include unified data source integration, API protocol field extensions for channel‑specific requirements, and template protocol matching for new activity page templates, all achieved by writing or updating functions without redeploying the underlying Java services.

Future directions involve building an IDE plugin for local function development, moving toward centralized function execution with shared middleware, expanding business scenarios, and promoting a Serverless‑for‑Frontend (SFF) architecture to reduce front‑back communication and improve overall development efficiency.