Bilibili DevOps Case Study: Culture, Community, User‑Driven Demand Management, High‑Performance Microservices, and Data Operations

1. Culture and History Bilibili (B站) originated as the Mikufans bullet‑screen community for "Hatsune Miku" fans, rebranded in 2010 to a video platform. It now hosts over 128 million monthly active users, with 78% aged 18‑35, forming a multi‑interest cultural community.

2. Community Operation Ecosystem The platform’s success stems from a strong community‑first culture, delivering high‑quality user‑generated content, interactive danmu, and a vibrant creator ecosystem. Bilibili’s annual New Year’s Gala exemplifies its deep understanding of Z‑generation user personas.

3. User‑Value‑Driven Demand Management Core user groups include anime fans, UGC creators, live‑stream enthusiasts, and general viewers. Demands are categorized into video content, user experience, community culture, and environment. Collection uses both business‑value analysis and real‑time feedback, with prioritization via a four‑quadrant method and commercial‑value weighting.

4. High‑Performance Microservice Practice Bilibili migrated from a monolithic PHP stack to a Go‑based microservice architecture. Key steps include vertical service decomposition (comments, coins, feeds, etc.), RPC framework built on Go’s net/rpc with Gob serialization, service registration/discovery (ZK‑based CP and polling‑based AP modes), and a centralized configuration center using MySQL and long‑polling. Performance optimizations cover link acceleration (DNS, CDN, protocol compression), a self‑developed gateway (dynamic config, routing, rate‑limiting, caching, auth, anti‑fraud), isolated deployment clusters, and robust reliability mechanisms (timeouts, multi‑level rate limiting, retry, circuit‑breaker, degradation strategies).

5. Data Operations The logging platform "Billions" (Go‑based) unifies collection, transport, splitting, and search, enabling trace‑ID based fault isolation. Monitoring consolidates Prometheus federated instances with remote storage in InfluxDB, providing end‑to‑end observability. The real‑time data platform "Saber" offers low‑code Flink jobs (BSQL, DAG) for feature engineering and AI integration.

6. Conclusion Over a decade, Bilibili’s user‑centric culture, agile DevOps processes, and robust data infrastructure have created a self‑reinforcing growth loop, positioning it for continued innovation despite not yet being profitable.