Data acquiring and processing from distributed terminal unit by network are usually real-time, large throughput and periodic, so data acquisition, processing and result should be done within data output period in high-efficiency. A real-time and large throughput data processing model using concurrency is built for this given application. The working principle and basic structure for the model are discussed in detail, also concurrency for the model and corresponding key buffer which concurrent threads put data into and gets data out are analyzed in particular. As is used in observation application networked by 1500 automatic weather stations, which has indicated that the model is capable in processing periodic surge observation data and has higher data throughput, as well as implements high speed and accurate results, high stability and reliability.