Constantly growing requirements for the information systems service quality require monitoring of network congestion. The experimental results proved that when using hybrid networks, it is difficult to estimate the network load based on the average time of information delivery due to the presence of a significant scatter of time data. A mathematical model of information transmission in wireless computer networks in the form of a queuing system has been developed. The system identifies the main elements: a data transmission channel, a data frame and a device that transmits data. Analysis of transmission standards allows us to identify the following main transmission modes for modeling: one frame successful transmission; transmitting a frame with an error and repeating the transmission (possibly several times); frame transmission using performance-enhancing technology (Bursting technology is used as an example). Events associated with the transmission of information over the wireless network of one device are considered as monitoring system states under consideration. Packet exchange process was considered to be a homogeneous Markov process with discrete states, continuous time and constant intensities. A mathematical model has been developed based on the state graph of the device-frame-transmission channel system, which allows one to determine the timing characteristics for the previously identified modes. The methods for assessing network congestion based on statistical information during data transmission is proposed. Delivery times are broken down into ranges derived from the simulation. The proportion of packets whose delivery time lies within the selected test range is estimated. To evaluate the results, experimental studies were carried out on a fragment of the hybrid network. Delivery time results for each experiment were divided into four ranges: corresponding to accelerated delivery of fragments, successful delivery on the first, second and third attempts, respectively. Results analysis showed that it is preferable to use the fourth range to assess network congestion. Approximation reliability in this case is 0.85, the correlation coefficient is 0.92. The proposed methodology allows us to estimate network congestion based on testing data with a fairly high degree of reliability.
Read full abstract