Enhancement of the graph neural network model for simulation modeling tasks

In the context of the rapid growth and increasing complexity of network systems, the task of accurately modeling network parameters becomes critically important for effective management and optimization of network infrastructure. In this context, the work focuses on an extended version of the RouteNet model, which enhances modeling capabilities by incorporating network node characteristics such as queue sizes. The original RouteNet model demonstrates good results in predicting delays and packet losses under homogeneous network conditions; however, it has significant limitations when it comes to accounting for variations in node characteristics. The extension of the model involves integrating mechanisms to handle nodes with different configurations, allowing the model to better reflect real network conditions. This includes considering node parameters such as queue sizes, which can vary significantly in real networks. This enhancement makes the model more adaptive and accurate in conditions that are closer to real network scenarios. The study includes a series of experiments involving the training and performance evaluation of the new model, its comparison with the original version, and an analysis of the results obtained. An important aspect is the assessment of the significance of these results for network research, particularly for tasks related to the simulation of computer networks. This not only improves the accuracy of modeling but also provides more effective solutions for managing and optimizing contemporary network systems, which is crucial for their stable and productive operation.