Abstract: Nowadays, the optical network is developing towards complex network environment and intelligent management and control system. There is no doubt that the intelligent management and control of the optical network is based on the accurate perception of the network. The digital twin technology uses a large number of perception units in the network system, combined with the network prior model, to carry out real-time and accurate “ mapping” of the network operating status and performance, thereby in the digital world an accurate “ copy” of the real network is obtained, enabling accurate analysis, automatic control and performance optimization of the real network. Data collection is an important part of the digital twin system in optical networks. Once a problem occurs during the collection stage, the performance of the system will be severely affected. This paper will conduct preliminary analysis and modeling of these problems, and discuss about the existing and potential solutions. 

Key words:  , all-optical network, digital twin, machine learning, perception of networks