Design And Implementation Of Functional Group Adaptation Mechanism Based On Traffic Awareness In Smart Collaborative Network

The traditional network has problems such as static and rigidity due to insufficient original design.There are serious drawbacks in terms of security and operating efficiency.Therefore,a series of researches on information network system and architecture have been carried out all over the world.As a typical representative,Smart Collaborative Network proposes a “three-layer” and “two-domain” system model which uses virtualization technologies to combine service function components into network functional groups as needed.It achieves diverse network services and meets the different needs of users.However,the foundation of deploying diverse network services lies in traffic awareness and adaptation,that is,classifying traffic features to redirect to the corresponding network functional groups.Traditional traffic awareness methods have gradually exposed the disadvantages of poor scalability to increasingly complex network environments.Machine learning uses traffic statistics to model traffic data from empirical data.It solves the problems of traditional methods and becomes the future development trend of traffic awareness.Based on machine learning,this thesis designs and implements a functional group adaptation mechanism based on traffic awareness in Smart Collaborative Network.This mechanism learns the traffic features in the network to realize traffic classification and redirects different traffic to the corresponding function groups for diverse services supply.The specific work is as follow:Firstly,based on the analysis of the current development and demand of the Internet,Smart Collaborative Network and machine learning are combined to perform the overall design of the functional group adaptation mechanism based on traffic awareness.The mechanism firstly obtains a variety of features including data packets and network flows by processing the original traffic;secondly,the mechanism uses machine learning to perform modeling and deploying with the extracted traffic features,so that it has realtime traffic classification capabilities;finally,the mechanism constructs the functional group and redirects the traffic to the corresponding network functional groups,which achieves dynamic adaptation between traffic and the functional groups.Secondly,this thesis specifically implements the functional group adaptation mechanism based on traffic awareness.Among them,the acquisition of traffic features is completed through three parts: packet capture,network flow processing and feature extraction.Traffic awareness work is divided into two parts: The offline learning part builds a learning model according to the supervised learning process and adopts the ensemble learning method Light GBM(Light Gradient Boosting Machine)as the classifier.The online awareness deploys the learning model online to provide real-time traffic classification.The mechanism constructs the network functional group through virtualization technology and sets a flow classifier to achieve redirection of traffic.Finally,this thesis uses the virtual network platform to build the system for verification and testing.Then the thesis verifies the functions and performance of each part of packet capture,network flow processing and feature extraction.Then it tests the feature processing methods involved in offline learning and uses a variety of machine learning algorithms to conduct traffic classification experiments.After that,the thesis evaluates the perception of different algorithms and chooses the best performing algorithm to test the overall mechanism.The system uses the obtained classification results to redirect different traffic to the corresponding network functional groups.Experimental result shows that the proposed mechanism can effectively use machine learning algorithm to establish models and perform dynamic functional group adaptation,which provides a basis and reference for the intelligent research of Smart Collaborative Network.