The problems of decrease in I/O speed and increase in data packet loss rate were caused by the increase of sampling frequency and concurrency of information data under the working condition of the silage harvester. Therefore, the data communication protocol and long string Codec methods of the vehicle terminals and cloud services were studied, the impact of Netty framework and traditional NIO framework on concurrency was analyzed, and three Codec methods such as Java serialization, Protobuf and Marshalling were compared and analyzed. The technical schemes of data collection, data transmission and Web application were put forward separately, and a remote monitoring system for working condition information of silage harvester was designed by use of CAN technology and Netty custom communication protocol. The simulation test showed that the system had a 08 times increase in concurrency, compared with the traditional data acquisition system, under the sending cycle of 500ms. At 200ms, 100ms, and 50ms transmission frequencies, the I/O speed of using Marshalling code was increased by 0.4 times, 3.9 times and 1.5 times compared with Java serialization, respectively. The 15 days of continuous field trials showed that the system run smoothly with good stability and reliability. The statistical analysis of the working condition data from the main components of the silage harvester showed that the data can be reference to the diagnosis of the working condition, and the system basically met the monitoring requirements of main components of silage harvester. The developed system can increase the I/O speed and ensure stable data access volume at high frequency and high concurrency. It had a great significance to promote the informatization and intelligent development of silage harvester mechanical equipment.