Service Governance Architecture and Practices at Zhuanzhuan

As companies scale, many move from monolithic to micro‑service architectures, splitting applications into independent services that are easier to manage, version, and evolve. However, micro‑services introduce complexity because inter‑service communication relies on RPC, requiring robust service governance such as registration, discovery, monitoring, authentication, and rate limiting.

The Zhuanzhuan Service Management Platform integrates service registration & discovery, a configuration center, monitoring, alerting, authentication, and rate limiting into a single solution that interacts with the RPC framework via an SDK.

When a service starts, it registers with the platform and subscribes to call relationships for authentication and throttling. Callers receive node up/down events and pull the latest node list via the SDK. Both callers and providers report latency, timeouts, exceptions, and distribution metrics to the platform, which also stores configuration parameters (e.g., timeouts, serialization protocols) that take effect in real time.

Regarding registration, the platform favors an AP model (e.g., Eureka, Nacos) to tolerate brief inconsistencies, using periodic SDK tasks to pull the latest node list and achieve eventual consistency. Nodes can be grouped to provide isolation for callers of different importance levels, and a gray‑release discovery feature allows selective exposure of specific service instances for staged rollouts.

The configuration center enables hot‑updating of RPC parameters without redeploying callers, supporting real‑time adjustments such as method‑level timeout changes.

Monitoring aggregates call metrics (total, average, max latency, percentiles, distribution) at the SDK level and performs multi‑stage aggregation before storage, reducing bandwidth and enabling millisecond‑level query performance.

Authentication and rate limiting are enforced using a unique methodKey format ( (${ServiceImpl})${ServiceInterface}.$method($parameterTypes)) that uniquely identifies RPC methods; the platform stores method lists uploaded by services and applies per‑service and per‑method policies via SDK filters.

Alerting integrates with the monitoring system to notify owners of exceptions, timeouts, or throttling events, with configurable intervals, methods, and target services.

Overall, the article provides a practical overview of Zhuanzhuan’s service governance architecture, highlighting its strengths and acknowledging remaining challenges such as high‑availability of notification mechanisms, consistency, and monitoring data storage.