How ZEEK’s Cloud‑Native Architecture Boosted App Stability and Agility

Background

Rapid growth of ZEEK digital services required a cloud‑native transformation to improve stability, iteration speed and issue resolution.

Guiding Principles

Adopt open‑source standards – use mature, widely adopted technologies (Java, Spring Cloud Alibaba) to avoid vendor lock‑in and simplify talent acquisition.

Leverage cloud value – delegate reliability, elasticity and component maintenance to the cloud provider, allowing engineers to focus on business logic.

Technical Practices

1. Unified Microservice Architecture

All business units migrated to a common stack: Java 17, Spring Cloud Alibaba, and Alibaba Cloud Microservice Engine (MSE). The framework provides integrated service registry (Nacos), configuration, and distributed tracing. This reduced inter‑team hand‑over cost and enabled a shared development baseline.

2. High‑Availability Service Deployment

Implemented loss‑less deployments using MSE cloud‑native gateway with the following mechanisms:

Readiness and liveness probes for graceful pod shutdown.

Pre‑warming of new pods before traffic switch.

Traffic‑aware auto‑scaling based on CPU, QPS and custom metrics.

Graceful deregistration from Nacos to avoid request loss.

3. Consolidated Traffic Gateway

Replaced separate traffic, API and microservice gateways with a single MSE cloud‑native gateway. Benefits:

≈50 % reduction in resource consumption.

Shorter request path and lower latency.

Built‑in WAF integration for DDoS and OWASP‑Top‑10 protection.

4. Serverless Visual Data Analysis Platform

Built the Jishu BI platform on Alibaba Cloud Serverless App Engine (SAE) Kubernetes. Key features:

One‑click deployment; container images built from source in docker build and pushed to ACR.

Namespace‑level multi‑tenant isolation.

Integrated logging, metrics and distributed tracing via SAE observability.

Supports horizontal pod autoscaling (HPA) and CronHPA for scheduled scaling.

5. Car‑Network Telemetry Ingestion

Data pipeline uses OSS as object storage and Function Compute (FC) for serverless processing:

Vehicle ECU encrypts payload and uploads to OSS.

OSS triggers an FC function.

FC validates, decrypts (using KMS), and writes cleaned records to Kafka.

Downstream microservices consume Kafka for analytics.

Serverless model provides automatic scaling to handle burst traffic and eliminates operational overhead.

6. AI‑Assisted Development (Tongyi Lingma)

Integrated Alibaba’s Tongyi Lingma large‑model platform via its REST API. Use cases:

Code skeleton generation from natural‑language prompts.

Custom command extensions for CI/CD pipelines.

Security‑compliant deployment inside VPC, with audit logs stored in CloudTrail.

7. Service Reliability Enhancements

Implemented the following patterns across microservices:

Graceful shutdown & readiness checks – ensure no in‑flight requests are dropped.

Horizontal sharding – large tables split across Polardb‑X instances; each microservice can scale independently across multiple AZs.

Traffic protection – rate limiting, MQ peak‑shaving, slow‑SQL throttling.

Circuit breaking – fallback to mock responses when downstream error ratio exceeds threshold.

8. Full‑Link Gray Release

Using MSE gateway traffic tagging, releases are rolled out by:

Deploying a gray version alongside the stable version.

Tagging incoming requests based on user segment or weight.

Gradually increasing traffic to the new version while monitoring health metrics.

Instant rollback by reverting tags.

This enables daytime releases without service interruption.

9. Unified Task Scheduling (SchedulerX)

SchedulerX replaces ad‑hoc cron jobs with a centralized service supporting:

Java SDK and Agent for custom tasks.

Compatibility with open‑source frameworks (XXL‑JOB, ElasticJob).

Second‑level scheduling and high‑availability across zones.

10. Automated Testing & Performance Validation

Function Compute is used to spin up isolated test environments that can:

Invoke health‑check endpoints after deployment.

Run regression suites against a snapshot of production data.

Perform load testing with configurable QPS and duration, feeding results back to auto‑scaling policies.

Future Outlook

The cloud‑native upgrades have already increased ZEEK APP stability and deployment agility. Ongoing work includes deeper observability, progressive migration to a service mesh, and expanding AI‑driven code quality checks.