Robust service quality prediction and guarantee method based on deep learning in SDN

Abstract

A robust service quality prediction and guarantee method based on deep learning in the SDN comprises the following steps of carrying out data preprocessing work on the collected switch port historicalflow data in a network topology, establishing a deep learning model, and training the deep learning model by using the preprocessed flow data; preprocessing the flow data of a switch port in a network topology collected by an SDN controller in real time, sending the preprocessed flow data into a trained deep learning model, and outputting QoS parameters of a current video flow service; calculating a QoE index according to the QoS parameter quantification of the front video stream service; and predicting, recording and monitoring the network traffic of the video stream service of the current user by using a machine learning method by using all QoS parameters of the current video stream service, and prompting a client or an SDN central controller to perform corresponding operation by usinga QoE index to realize a traffic prediction function.

Description

technical field [0001] The invention belongs to the control layer technology in a software-defined network, and specifically relates to a robust service quality prediction and guarantee method based on deep learning in SDN. Background technique [0002] With the continuous development of my country's information and communication technology, the emergence of 5G marks another innovation in mobile communication technology. As a key technology in the development of 5G, the emergence of software-defined networking (SDN) separates the control plane and data plane in traditional networks, making it easier to obtain and identify network traffic-related data. It greatly simplifies the control and management of the network layer, promotes the speed of network deployment in the past, and enables centralized control of traditional networks that are intricately connected. At the same time, for service providers and operators, the emergence of software-defined networks also provides the...

AI Technical Summary

Problems solved by technology

[0008] First of all, traditional methods cannot handle a large number of network eigenvalues ​​very well. Once there are too many or too few eigenvalues, it is easy to overfit or underfit, and the prediction accuracy will be affected.

For example, the random forest method used by the Royal Swedish Institute of Technology mentioned above, for data with attributes with different values, attributes with more value divisions will have a greater impact on the random forest, so the random forest produced on this data What is lacking in the credibility of attribute weights

[0009] Secondly, the traditional QoS prediction method has a serious lack of prediction accuracy for noisy data. However, network data is often intricate, and data noise is difficult to avoid in practical applications. Traditional QoS methods cannot provide high prediction accuracy.

At the same time, in terms of real-time forecasting, the forecasting speed of traditional methods is slow, and real-time performance is difficult to guarantee

[0010] Finally, some traditional prediction methods need probes to detect the content of user data packets to analyze the service quality of terminal applications, such as using DPI technology, etc. This method not only cannot handle encrypted data packets, but also is not conducive to the protection of the privacy of ordinary users

At the same time, the traditional QoS prediction is only the prediction of network machine parameters, without considering the specific human-centered perception, that is, the parameter problem of QoE

Images