OceanBase has publicly revealed its data architecture practices supporting the AI generation application of Ant Lingguang for the first time. Through technologies such as logical tables for massive applications, shared physical storage, and controlled SQL computing, it achieves a new database architecture that enables "independent application data models and shared underlying resources," providing a data foundation for large-scale operation of AI generation applications.
Data from April this year shows that the Ant Lingguang platform has generated approximately 30 million flash applications in total. With AI lowering the barriers to application development, a large number of applications generated by Agents have grown rapidly. Traditional database architectures now face new challenges: they need to allow each application to have independent data management and computing capabilities while avoiding the storage and maintenance pressures caused by building databases separately for massive applications.
To address this issue, OceanBase designed a "logical table" solution tailored for massive AI applications. Each flash application still has an independent data table structure at the usage level, allowing it to create tables, write data, and perform standard SQL queries. However, the underlying layer shares a unified physical storage. By integrating with the OceanBase JSON Table SDK, application data is stored in shared tables in JSON format. The database completes data mapping, querying, and aggregation calculations based on the application's independent schema, achieving precise data processing capabilities.
Compared to traditional solutions that store data centrally in JSON large tables or create separate physical tables for each application, this architecture reduces the database management pressure caused by a large number of applications while ensuring multi-tenant permission isolation. When a single application scales up, it can be migrated to a separate physical table to achieve higher performance.
OceanBase stated that this practice not only verifies the data management capability of AI-generated applications but also reflects the trend of databases evolving into AI data platforms. In June this year, OceanBase released its next-generation lakehouse-integrated AI database, which unifies the management of structured, unstructured, and vector data, supporting AI Agents to obtain complete business context. It has been verified in scenarios such as Ant Afu and Lingguang, and is advancing enterprise customer testing.
As AI Agents gradually become the next generation of application forms, the importance of data infrastructure continues to rise. OceanBase aims to build a data foundation tailored for Agents to provide more stable data support for the large-scale deployment of enterprise AI applications.
