A distributed data governance architecture for smart cities based on data unified registration, dynamic catalog, and local LLM data coordination

Abstract

Effective governance and secure sharing of data resources remain key challenges in smart-city development, where cross-department collaboration is constrained by data ownership boundaries and compliance requirements. This paper proposes a distributed data governance architecture that integrates a unified data registration standard, a probe-driven dynamic data catalog, and on-premises local LLM data coordination to support compliant cross-system discovery and access. The architecture follows a “one body with two wings” ownership-centric approach that balances lawful sharing and security protection, and it explicitly bounds the Local LLM to semantic interpretation and explanation while deterministic policy engines and human review enforce final decisions. We report a production deployment in Suining City, China, spanning 50 departments and 151 systems. A representative load test on the deployed platform processed 83,909 simulated requests and achieved an average response time of 730 ms, a maximum latency of 1,443 ms, and an observed error rate of 0.12%, demonstrating stable city-scale operation. Overall, the work provides deployment-grounded architectural insights for scalable and policy-pluggable smart-city data governance.