Методика получения обобщенных статистических характеристик сетевого оборудования для создания имитационных моделей

Abstract

The research is devoted to the problem of the development models of network equipment that precisely reproduce the operation of real equipment by the time characteristics. Aim. This research is aimed at developing a methodology that allows you to quickly obtain models that statistically coincide in the behavior of the selected parameters with real measured equipment on existing open modeling systems. Methods. To obtain the performance parameters and time characteristics of the real equipment, an approach was proposed, which includes experiment planning to compute the characteristics for a wide range of input traffic intensities. In addition, the approach describes the connection scheme and equipment configuration including external switch control to test the effect of various settings on the performance and packet processing times. It also includes the scripts to automate the creation of traffic test sets with the required parameters of time distribution between the packets as well as the scheme of connection through the container services and receiving traffic dumps. The approach does not include the testing of the output queues operation because the goal was to investigate and transfer into the model processes, which occur before the frames are placed in the outgoing interface queues. An open-source Omnet++ software was used as a modeling environment. Results. We studied the structure of its existing switch model and identified disadvantages that affect the accuracy and reliability of the results. In addition, we proposed an approach for implementing model delay lines based on the interpolation of time parameter tables obtained from real data for the required combination of traffic parameters and inbound interfaces. To test the developed approaches, a test bench on real equipment was assembled, and an experimental research plan was developed. Finally, an automated testing script was implemented according to a given plan. Testing was carried out in the conventional frame switching mode as well as in the switching mode based on OpenFlow tables. In addition, the time characteristics of the intermediate equipment were also obtained in order to take into account their influence on the experimental results. Based on them, a table of parameters was computed for the processing delay line in the model, and the experiment was conducted on the model in Omnet++. Since the obtained samples in the statistical analysis showed the types of distributions that are different from the 3 normal and exponential, the statistical Mann–Whitney U-test was used to compare the mean values. Conclusion. The results obtained at a significance level of 0.05 confirmed the efficiency of this approach for models that more accurately describe the temporal characteristics of the traffic processing in network equipment.