Meeting the Artificial Intelligence Challenge: Building the Next-Generation DBMS

With the explosive development of generative artificial intelligence (AI), data management is confronted with new challenges such as explosive data growth, real-time dynamic updates, multi-modal semantic fragmentation, and privacy security, making traditional database systems incompatible with the needs of the AI era. This article analyzes the core key technologies of database management systems driven by AI, including issues like ultra-large-scale unstructured data processing, semantic fusion of multi-modal data, real-time update efficiency, and privacy protection. Against this backdrop, Shenzhen Institute of Computing Science has adopted a “theory + engineering” dual-track model, drawing on the R&D paradigm of Bell Labs, to build the fully self-developed Yashan Database System (YashanDB). The system has achieved high-concurrency transaction processing, scale-independent query processing, real-time computing of big data, unified management of multi-modal data, and data security and privacy protection through key technologies such as the original runtime scheduling method based on transaction cost, fine-grained multi-version concurrency control, bounded resource computing theory, shared cluster architecture, including cohesive memory and decentralized transaction management, multi-modal semantic connection, and lightweight privacy enhancement. It effectively addresses the data management challenges in the AI era. YashanDB has been applied in core systems of key fields such as finance, energy, and government affairs, providing a self-controllable core data base for the AI era.