How Vivo Scaled Its Microservice Platform for 500M Users: Architecture & Lessons

Background and Motivation

Rapid business growth at Vivo led to increasing service complexity, user scale, and traffic volume. To meet higher delivery speed, scalability, and reliability requirements, the company decided to adopt a microservice architecture.

Why Microservices and Initial Challenges

Key pain points included low efficiency in configuration changes and service releases, difficulty in high‑availability assurance and fault isolation, performance bottlenecks in cross‑service communication, and high collaboration costs among upstream and downstream teams.

Architecture Capability Matrix

Vivo identified three core business scenarios for microservices: synchronous RPC calls, asynchronous messaging, and scheduled tasks. The required technical capabilities were:

Synchronous calls: RPC framework, service registry, service governance.

Asynchronous calls: Message middleware.

Scheduled tasks: Distributed task scheduler.

An ability matrix covering roughly 30 items was created, organized across the access layer, service layer, and data layer, and aligned with DevOps stages such as development, operation, and monitoring.

Platform Capabilities

The platform consists of five layers:

Access layer: Four‑layer traffic gateway and API gateway.

Service layer: Service/traffic governance platform, configuration center, registration center, API management, distributed task scheduling.

Message layer: Message middleware.

Framework layer: Scaffolding with SDKs for logging, configuration, rate‑limiting, circuit‑breaking, MySQL/Redis, and RPC.

Storage layer: DaaS platform providing MySQL, Redis, Elasticsearch, MongoDB, and file services.

Observability is supported by a monitoring center, log center, and tracing system. CI/CD and CMDB are also integrated.

Engine Upgrade – Registration Center

Problems with the original ZooKeeper‑based registration center included CP constraints, lack of multi‑datacenter active‑active capability, limited write scalability, and high fault‑radius. Vivo upgraded to an application‑level service discovery model, reducing memory usage by 50 % per node and cutting storage/push pressure by 90 %.

The new design separates Session and Data layers: Session handles long‑lived client connections, while Data persists registration information. This architecture provides AP characteristics, cross‑datacenter active‑active support, and horizontal scalability. The internal project code‑named vns implements this solution.

To avoid tight coupling, vns exposes both gRPC and ZooKeeper protocols, allowing a gradual migration and gray‑scale rollout. If vns is not yet production‑ready, a temporary ZooKeeper cluster can be used, later replaced once vns matures.

Engine Upgrade – Configuration Center

The legacy configuration center suffered from fragmented change channels, slow rollback, scattered audit logs, and insufficient performance. Vivo built a unified configuration channel supporting both application‑level and component‑level configs, adding one‑click approval, audit, and rollback capabilities.

The upgraded system improves high‑availability, performance, security, and observability, while maintaining backward compatibility to ensure a seamless migration.

Unified Microservice Platform Construction

Vivo identified the configuration center and registration center as the foundational components for a unified platform. By consolidating high‑frequency functions from multiple modules, a single entry point was created, reducing duplication and operational overhead.

Key improvements include:

Unified configuration and CI/CD integration for one‑click approval and rollback.

Platform‑wide automation to boost continuous delivery capabilities.

Middleware Component Lifecycle Management

Vivo established a full lifecycle management strategy for middleware components, covering demand analysis, component scanning, standardization, version governance, and maturity assessment using a Gartner‑style matrix. The approach balances rapid iteration with strict quality controls.

Future Directions

Vivo plans to continue evolving the platform with cloud‑native technologies such as Service Mesh, Serverless, and compute‑storage separation. The goal is to enhance development efficiency, reduce resource costs, and support emerging business scenarios like mini‑programs and fast‑app ecosystems.

Overall, the microservice platform demonstrates how a large‑scale internet company can combine open‑source components with selective self‑development to achieve high scalability, reliability, and operational efficiency.