Abstract: WMO Information System (WIS) is a coordinated, distributed, global infrastructure for the collection and sharing of information for all WMO and related international programs. As the core center of WIS, each Global Information System Center (GISC) is responsible for the collection and distribution of global exchanging data, and providing data discovery and access service. As a GISC of WIS, a scalable and flexible system is designed and established to satisfy WIS/GISC functionalities in Beijing. Beijing Global Information System Center should hold at least 24-hour WMO global exchanging data files, which could be accessed by authorized users through DAR (data discovery, access and retrieval) services.GISC Beijing has to do validation check for all the global exchanging data files, and only files matching the corresponding metadata could be brought into data cache. The existing approach for the validation is based on database retrieval operation. Currently, there are more than 100000 metadata records stored in the relational database GISC Beijing, while the system receives more than 50000 global exchanging data files in an uneven distribution of collection time. The disadvantage of this way is that frequent database I/O operation would lead to a sharp decline of the system performance, especially when a large number of data entering. Therefore, the approach could not satisfy the requirement of real-time data cache service. Although establishing the table index and multi-threaded mechanism could solve the efficiency of data processing to some extent, it is inevitable that frequent database I/O operation would bring about the performance bottleneck. Therefore, the operation treatment should be optimized.It is a possible way making full use of memory technology to reduce disk I/O cache data and improve the efficiency significantly. Considering the complexity of the mature memory database, a more targeted approach is adopted which is suitable for the scenario of dynamic nature of data with the timeliness requirements. An application is designed and implemented based on the memory object caching technology. When initializing the application, the system loads metadata into memory as a hash table from the database based on stored key/value pairs, organized by the unique bulletin head information of global exchanging data. In this way, the metadata contents are encapsulated as memory objects, thereby providing fast data memory retrieval method. In addition, parallel processing is implemented to extend the functionalities, including data cache logging function and data subscription services.As a result, effects of the optimized function can satisfy the real-time business requirements, reducing the data processing time to an average of less than 5 ms. It also provides an easier way to do extensions by adding memory object using parallel processing.