Logistic Regression-based Solution to Predict the Transport Assistant Placement in SDN networks
Luis Jesús Martín León, Juan Luis Herrera, Javier Berrocal, Jaime Galán–Jiménez · 2023
During the last years, applications requirements have been changing and the Information and Communication Technology (ICT) sector had to evolve and explore new solutions to approach the requirements of applications of the future such as telepresence, augmented reality, metaverse, holoportation, etc. One of the aspects that could serve as a basis to satisfy the stringent QoS requirements of the applications of the future is the reduction of the latency caused by TCP retransmissions. Through the proactive location of a novel network function, namely Transport Assistant (TA), the delay caused by TCP retransmissions is reduced, thus improving the network QoS and satisfying the QoS required by the applications. In this paper, a Machine Learning solution based on Logistic Regression (LR) is proposed to predict which is the part of the network that is prone to negatively impact the network performance. Through experiments based on the training of historical data, the LR-based solution is able to predict the correct location of the TA with a precision average of 95% and accuracy of 90%. Such good results in the prediction help to make better decisions and therefore to save time and resources, improving the network management.