The successful functioning of telecommunication networks largely depends on the effectiveness of algorithms for detection andprotection against overloads. The article describes the main differences that arise when forecasting, monitoring and managing congestion at the node levelandat the channel level. An algorithm for detecting congestion by estimating the entropy of time distributions of traffic parameters is proposed.The entropy measures of data sets for various types of model distribution, in particular for the Pareto distribution, which optimally describes the behaviorof self-similar random processes, were calculated and analyzed. The advantages of this approach include scalability, sensitivity to changes in distributionsof traffic characteristics and ease of implementation and accessible interpretation.
