How SKG Built a Scalable Cloud‑Native Channel Platform with Alibaba SAE

Project Background

SKG, a manufacturer of high‑end health products, needed a unified channel middle‑platform to resolve inaccurate inventory, poor online‑offline coordination, inflexible deployment architecture, and slow IT response. The goal was to support rapid sales expansion and scalable operations by moving to a cloud‑native solution.

Existing Architecture and Pain Points

Before the migration, SKG operated a hybrid environment consisting of an IDC data‑center, Alibaba Cloud ECS instances, and RDS. Application release, open‑source integration, cloud service binding, and cluster management were all performed manually, leading to:

Low DevOps maturity : Manual deployments, long release cycles, high risk of production failures.

Complex deployment workflow : Resource evaluation, server provisioning, software installation, plus separate components for monitoring, scripting, configuration, and logging.

High learning curve for containerization : Developers unfamiliar with Kubernetes details.

Inconvenient elasticity : Under‑utilized resources during low demand and cumbersome scaling procedures.

High operational cost : Ongoing maintenance of both applications and underlying infrastructure required significant manpower.

Technical Requirements

Unified online delivery platform with full lifecycle control, CI/CD automation, and quality assurance.

Seamless compatibility with a Spring Cloud‑based microservice architecture.

Operation‑free environment for developers, allowing focus on business logic.

Flexible elastic scaling and resource optimization.

Comprehensive microservice support (gray release, traffic control, remote debugging, monitoring, etc.).

Technology Selection and Comparison

The team evaluated three options: pure ECS, self‑managed Kubernetes, and Alibaba Cloud Serverless Application Engine (SAE). SAE was chosen because it provides a managed container runtime with serverless characteristics, eliminating the need to manage underlying ECS or K8s clusters.

Key differentiators of SAE over the other options include:

Lower operational overhead – the platform handles cluster provisioning, scaling, and health‑check management.

Instant provisioning – containers can be launched in seconds rather than minutes/hours.

Built‑in DevOps features – native CI/CD pipeline integration, automatic image rollout, and version control.

Native integration with Alibaba Cloud services such as ACM (Configuration Management) and ARMS (Application Real‑time Monitoring).

Implementation Details

Deployment was performed through Alibaba Cloud’s Feitian technology service platform – Dayu – which provides a unified CI/CD pipeline and governance console.

Application packaging : Each microservice (Spring Boot, Spring Cloud, Dubbo, Node.js, etc.) was containerized using Docker, pushed to a private image repository, and referenced in the SAE console.

CI/CD pipeline : Dayu’s pipeline automatically builds the Docker image, runs unit tests, and triggers a deployment to SAE via the sae deploy API. The pipeline definition is stored as code, enabling repeatable releases.

Configuration management : Global configuration items were stored in ACM and referenced by the services at runtime, eliminating manual config file edits.

Monitoring and tracing : ARMS agents were injected into the containers to collect latency, error rates, and resource usage. Metrics are visualized in a single ARMS dashboard.

Elastic scaling policies : SAE scaling rules were defined based on CPU and memory thresholds. When a threshold is crossed, SAE automatically adds or removes container instances without user intervention.

Results and Metrics

Deployment speed : 20+ applications were initialized, created, and deployed to SAE within 2–3 hours.

Cost reduction : Resource cost decreased by >30 % compared with dedicated ECS instances. Test environments using 0.5‑core specifications saved >50 % of resource consumption.

Scaling efficiency : Scaling time improved from days (manual ECS/K8s) to minutes (SAE auto‑scale).

CI/CD efficiency : Over 3 000 automated pipeline releases were executed, with an average release time of ~100 seconds and zero release failures, representing a 300 % increase in deployment efficiency.

Product Improvement Suggestions

Expose more OpenAPI endpoints for synchronizing project data (requirements, tasks, defects, man‑hours, documents).

Enhance microservice governance features such as online debugging, service mocking, and consumer‑view testing for Feign, Dubbo, and HSF.

Provide a global monitoring view that aggregates metrics across all microservices rather than per‑application dashboards.

Improve SLB integration handling to automatically clean up stale entries after load balancer deletion or failure.

Add dynamic hot‑deployment capabilities to further accelerate iteration cycles.

Allow NAS storage credentials to be specified in container images to enable secure NAS access during runtime.

Conclusion

By migrating the channel middle‑platform to Alibaba Cloud SAE and using the Dayu delivery platform, SKG achieved rapid, low‑cost, and scalable deployment of more than 20 microservices. The serverless container approach eliminated most operational overhead, provided built‑in DevOps automation, and delivered elastic resource management, positioning the company for continued digital transformation and market expansion.