OceanBase: Practices of a Distributed Relational Database in the Internet Era

At the 2017 China Big Data Technology Conference (BDTC) in Beijing, Ant Financial senior researcher Yang Zhenkun delivered a talk titled “OceanBase—Relational Database Practices in the Internet Era,” outlining the current state and pain points of relational databases, such as high cost, limited scalability, and consistency challenges.

He explained that traditional primary‑secondary replication suffers from a trade‑off between consistency and availability, especially under heavy load like Double‑Eleven shopping festivals, and that replacing databases is risky and costly.

OceanBase addresses these issues by providing a distributed architecture that offers elastic scaling, ten‑fold cost advantages, and high performance without heavy reliance on disk I/O, using multi‑active primary‑backup setups and majority‑based transaction logging.

The system employs automatic leader election across multiple zones, avoiding single points of failure; a typical deployment uses one primary and multiple backups (e.g., one‑primary‑two‑backup or one‑primary‑four‑backup) with Paxos‑style consensus to ensure transaction durability.

Additional features include read‑only replicas for high read traffic, failover databases for emergency routing, and off‑peak data compression to reduce storage costs while maintaining performance.

Since 2013, Ant Financial has progressively migrated core systems to OceanBase, achieving high utilization rates and cross‑city deployments, and has begun offering the platform to external financial institutions.

The presentation concludes with reflections on the ongoing challenges of database reliability and future optimization directions.