Medindo eficientemente a vazão em redes de banda larga: uma abordagem baseada em aprendizado de máquina

Abstract

Network monitoring is essential to determine the quality of broadband services offered by ISPs. Establishing which metrics are relevant for this type of evaluation still remains an object of study. One of the main metrics of interest is the measurement of throughput between two endpoints in the network. The current most used approach for this type of measurement is inefficient and transmits a huge amount of data. This work proposes a new method that combines the transmission of small bursts of packets and a K-Means filter to calculate the available throughput, reducing the amount of data and time spent. Simulations and real experiments were conducted and show good results when compared to the common approach.