Implementing a Microservice Architecture with Spring Cloud, Docker, and a PaaS Platform

The goal of the PaaS cloud platform built on microservice architecture and Docker containers is to provide developers with a fast development, deployment, and continuous integration workflow, offering infrastructure, middleware, data services, and cloud servers so that developers only need to submit business code and minimal configuration.

Microservice access follows the path: external request → load balancer → service gateway (GateWay) → microservice → data/message services. Both the gateway and microservices rely on service registration and discovery, while all clusters obtain configuration from a configuration center.

Implementing microservices requires substantial infrastructure effort, but open‑source frameworks such as Netflix, Spring Cloud, and Dubbo alleviate this. The article focuses on Spring Cloud, especially the Spring Cloud Netflix suite, which provides many of the needed components.

An architecture diagram (see image) illustrates the request flow and shows that the gateway routes requests dynamically, performs rate limiting, authentication, monitoring, and logging.

The service gateway is deployed as a cluster behind Amazon EC2 instances with Elastic Load Balancing (ELB) for load distribution and Nginx for SSL termination. Its core capabilities include dynamic routing, throttling and fault tolerance, identity authentication, monitoring, and access‑log collection.

Spring Cloud Netflix’s Zuul is used to implement the gateway, leveraging a series of customizable filters to realize these functions.

Service registration and discovery are handled by Eureka. Each microservice registers with Eureka and sends periodic heartbeats; the server removes instances after missed heartbeats unless self‑protection mode is active. Eureka nodes synchronize via DNS‑based clustering.

Microservices are packaged as Docker images and run in Docker containers managed by Docker Swarm. An image repository stores base and final images, simplifying deployment and scaling.

Fault tolerance is achieved with Netflix Hystrix, which provides circuit breaking, thread‑pool isolation, fallback mechanisms, and rate limiting to protect the system from cascading failures.

A dynamic configuration center is built with Spring Cloud Config Server. Configuration files reside in a private Git repository; when changes are pushed, a server‑side hook notifies the config server via a message queue, allowing services to refresh configurations without redeployment.

The article concludes by noting that additional components such as logging and messaging services may be added as needed, and provides a reference link to the original Chinese blog post.