How HeLian Health Scaled Its Low‑Fault Medical Platform with Cloud‑Native Microservices

Technical Evolution Stages

HeLian Health, founded in 2014, provides SaaS solutions for pre‑, during‑, and post‑medical examinations, serving hospitals, enterprises, and families across more than 200 cities and 2,000 hospitals.

Stage 1 – Macro Application : Rapid 0‑to‑1 development using an early Alibaba Cloud container‑management service. Speed was prioritized, leading to significant technical debt, frequent failures, and unmet business expectations.

Stage 2 – Initial Microservices : As hospital integrations grew, failures increased. The team introduced Dubbo and Nacos for modular decoupling, but insufficient business understanding caused excessive cross‑service calls and a “super service” that harmed stability.

Stage 3 – Microservice Refactor : Focused on order processing, data synchronization, and a full migration to Kubernetes. Middleware was replaced with Alibaba Cloud Microservice Engine (MSE), resulting in a more stable system and improved development‑operations efficiency.

Low‑Fault Tolerance Challenges

The medical check‑up business is inherently low‑fault tolerant: any IT issue that prevents a user from completing an appointment severely damages experience. The domain also suffers from low traffic (1‑2 exams per person per year), making gray‑release ineffective and allowing bugs to surface months after deployment.

Service‑Oriented Architecture Design

To address complex business logic, HeLian modularized and decoupled services. Two splitting modes are used: business‑centric and capability‑centric. The capability‑centric split (orders, payments, messaging) forms the core, while peripheral services such as hospital data sync and manual fulfillment operate as independent modules.

Business‑critical services are isolated from stable services via HTTP gateways; within the cluster, Dubbo handles RPC, Nacos provides service registry and configuration, and RocketMQ manages asynchronous messaging.

Dubbo + Nacos Stack

Dubbo, a Java‑based RPC framework, requires only simple annotations (e.g., @DubboReference) to expose services. Nacos integrates seamlessly, offering an easy‑to‑use control panel. Early on, the self‑hosted Nacos community edition faced performance bottlenecks under high service counts.

Switching to Alibaba Cloud MSE alleviated pressure, delivering 4× performance gains and lower resource costs. Upgrading from Dubbo 2 to Dubbo 3 resolved the one‑interface‑one‑service scalability issue.

Feature Flags and Security

MSE’s dynamic configuration allows safe toggling of feature flags without restarts, and integrates with Alibaba Cloud KMS for encrypted storage, keeping users unaware of the underlying changes.

DevOps Practices

HeLian deploys containers on Kubernetes (ACK) with image‑based rollbacks. Continuous integration uses Alibaba Cloud CloudEffect CI, enabling 20‑30 releases per week and reducing deployment time from 2‑3 hours to under 8 minutes.

Service isolation is achieved by deploying multiple versions of the same service side‑by‑side, each with independent memory; a failure in one version does not affect the other.

Future Roadmap: Service Mesh

The next goal is a Service Mesh control plane (Alibaba Cloud ASM based on Istio) to enable full‑link routing, version‑aware traffic splitting, and reduced test‑environment costs. By tagging traffic at the gateway and propagating tags across service hops, each team can test in isolated environments without maintaining multiple full stacks.

Long‑term, HeLian plans to replace traditional web layers with a BFF (Backend‑for‑Frontend) gateway, exposing backend services directly to the front‑end as the industry shifts toward heavier front‑end responsibilities.

References

https://www.aliyun.com/product/aliware/mse

https://www.aliyun.com/product/kubernetes

https://help.aliyun.com/document_detail/214749.html