Analysis of internet network performance in the territory of the Russian Federation using data from the RIPE Atlas distributed measurement system

This article is about analysis and evaluation of the Internet performance in the country. A set of different approaches to estimation of the network performance is proposed, including working with anomalous data, comparing different territorial zones and analyzing the components in the obtained time series. The paper focuses on factors affecting the Internet connection speed, such as distance to servers, time and phase of the day and also the provider, which allows identifying the main factors of network performance. In addition, methods for categorizing time series are considered, namely machine learning models for clustering network traffic data. This allows not only to analyze the current situation, but also to more accurately predict future network performance values. The results can also be applied in adaptive traffic management, which can improve the efficiency of using network resources.