Microservice Gateway (Kong) and Container Service Integration

The 360 private‑cloud container service team collaborated with the AI research institute to enhance machine‑learning efficiency in the cloud, delivering services such as face detection, screen‑artifact detection, porn detection, pet detection, image stylization, OCR, and smart cropping. The article then focuses on the "Microservice Gateway and Container Service" technologies used in this practice.

Microservice architecture breaks a monolithic application into many small, independent components that can be developed, deployed, and scaled separately, possibly using different languages and teams. Each component collaborates to accomplish the overall task, embodying the principles of being smaller, faster, and stronger.

In contrast, traditional monolithic web applications (e.g., a Java WAR file) are packaged as a single unit, which simplifies testing and deployment but limits scalability, reliability, iterative development, multi‑language support, and team collaboration.

To address these drawbacks, the monolith can be split into independent services, allowing each team to choose its own technology stack and iterate without affecting the whole system.

An API gateway provides a unified entry point for external clients to access one or more internal APIs, offering features such as load balancing, rate limiting, authentication, monitoring, and caching. It hides internal services while exposing only the necessary APIs, and can enforce security and stability policies.

Kong is a high‑availability, extensible API gateway built on the nginx_lua module. It consists of a Kong server (based on nginx) that receives external API requests and a PostgreSQL database for storing configuration data. Kong’s plugin mechanism, written in Lua, allows custom functionality to be executed during the request‑response lifecycle.

The integration example shows how three containerized AI services—face detection, pet detection, and content‑moderation—are deployed on the container platform, each receiving a VIP address. These services are then registered with Kong, which persists the registration in PostgreSQL and configures plugins for authentication, rate limiting, CORS, etc.

Step 1: Deploy the three service instances on the container platform and obtain VIPs. Step 2: Register the services with Kong; Kong stores the information in PostgreSQL and applies necessary plugins. Step 3: Clients obtain an authorized token and access the desired services through Kong.

The article provides a high‑level overview of the workflow; each part requires further in‑depth study by the reader.