Abstract: The enterprise architecture (EA) reflects its business components and process, application system, data composition, technical composition and organization setting, so it's an important tool to realize the strategic goal of the enterprise. EA includes frameworks and reference models, consisting of business architecture and IT architecture.China Integrated Meteorological Information Sharing System (CIMISS) is the unified data environment which is distributed in national service center and 31 provincial services centers. Main functions of CIMISS include data collection and dissemination, quality control and product generation, data management, data services and operational monitoring, which achieves the initial unity of data standards and the standardized management of meteorological data. CIMISS provides direct data services to meteorological applications, and is put into operation at the end of 2016. CIMISS obtains good efficiency of overall data process and service, the collection and storage time of core data is less than 3 minutes, the data access performance is enhanced by 2-5 times compared to the national meteorological archival and retrieval system (MDSS).The architecture design is an important stage of the construction of CIMISS which gives strong foundational support to the detailed design and development of the application software system. The architecture design of CIMISS and its core components are introduced and illustrated, combined with the case analysis of the application practice.Applying EA methodology, in the form of top-down and layer-by-layer decomposition, the business architecture, the data architecture, the application architecture, the technique architecture and the standardization framework are established. By adopting a series of optimized architecture design, the effective management and synchronization of metadata and data quality control are realized, the scalability of unified monitoring of new business and the rapid online service of new meteorological data are provided, the universal data service model of heterogeneous database systems is established, the good performance of data services is achieved and is easy for the upgrade of the database technology. Through the hierarchical design and the application of cluster technology, requirements of maintainability, scalability and high reliability of CIMISS are satisfied.Based on CIMISS, the meteorological data service platform under design can be upgraded to be compatible with new technologies such as cloud computing and big data. The intensive infrastructure service platform would be built, and distributed data curation technology would be implemented comprehensively in data management to achieve high performance. Through application programming interfaces (API), operational applications wouldn't be affected by the update.