Vivo 2022 Dev Conference: Frontend Compiler, Low‑Code, Real‑Time & Cloud‑Native

#1 Frontend Innovation: wepy‑chameleon Compiler

Vivo engineers presented the wepy‑chameleon compiler, which converts legacy wepy mini‑program code to the more widely adopted uniapp framework with a single click, addressing the declining community activity of wepy and easing migration to modern cross‑platform solutions.

#2 Low‑Code Platforms: Houyi and Game Center

The Houyi low‑code platform enables rapid visual development of backend B2B systems with a front‑back separation architecture, offering flexibility, extensibility, and high integration. A separate low‑code solution for the game center provides component‑based, dynamically adjustable UI construction, improving development speed and product quality.

#3 Real‑Time Computing Platform

Vivo's real‑time computing platform, launched in 2019 and built on Apache Flink, now supports SQL capabilities, containerization, and efficiency‑focused features such as unified batch‑stream processing and task diagnostics, serving data‑driven services like real‑time warehousing, recommendation, advertising, security, and monitoring.

#4 Branch‑Path Graph Scheduling Framework

To overcome limitations of traditional DAG scheduling in Vivo's effect‑advertising recommendation system, a new graph scheduler adds branch nodes and supports both "AND" and "OR" activation, enabling more flexible path selection and improving system extensibility and entropy reduction.

#5 Heterogeneous Mixing for Ads

Vivo's mixing engine balances user experience with revenue by integrating heterogeneous content queues, handling differing modeling goals, extensive product rule constraints, and preserving item order, with three iterations: fixed‑position mixing, Q‑learning based mixing, and deep‑learning position‑aware mixing.

#6 Cloud‑Native & Container Practices

Since 2018, Vivo has built a one‑stop cloud‑native machine‑learning platform on containers, supporting data management, model training, and deployment. Recent efforts focus on high‑availability Kubernetes clusters, automated operations, and continuous upgrades from v1.10 to v1.17, leveraging new controller features for stability and efficiency.

#7 Monitoring System Evolution

Vivo's monitoring platform now unifies infrastructure, platform, and custom monitoring services, covering IaaS, PaaS, DaaS, and CaaS, and transitioning from DevOps to AIOps to handle increasingly complex, heterogeneous service graphs.

#8 Database Service Platform (DaaS)

The self‑developed DaaS platform automates database operation, offering intelligent resource allocation, elastic scaling, isolation, and transparent MySQL encryption for personal data, thereby improving efficiency, reducing management cost, and ensuring data security.

#9 Messaging Middleware: RocketMQ, Kafka, Pulsar

Vivo employs RocketMQ for online services and Kafka for high‑throughput data ingestion, sharing best practices on resource isolation, traffic balancing, rate limiting, and cluster governance. Future plans include transitioning to Pulsar to further enhance performance and functionality.