Design And Implementation Of Probe Selection System Based On Active Measurement In Multi Domain Network

Network measurement can monitor network performance and model network space,which is the basis for ensuring network security and quality.In order to measure the network,the network administrator needs to select appropriate nodes in the cross-domain and heterogeneous network to deploy the measurement nodes,so that the consumption of the entire network is minimized and the impact on the network performance is minimized.With the rapid development of network technology,the scale and complexity of the existing network are increasing day by day,which brings great difficulties to network measurement.It is very important to design a suitable measurement node selection algorithm.Most of the current measurement node selection algorithms are based on genetic or heuristic algorithms to construct the minimum weak vertex cover problem and model the network topology.Such algorithms do not take into account the heterogeneity of real networks,nor do they take into account the multi-autonomous domain situation in real networks.In addition,there are many evaluation indicators for the selection of measurement nodes,and no unified and complete evaluation standard has been formed.Aiming at the problems existing in the selection of measurement nodes,this thesis makes the following contributionsThis thesis proposes a measurement node selection algorithm that can be applied in cross-domain and heterogeneous networksHeterogeneous network node selection(HNNS)algorithm.The algorithm first uses the heterogeneous network representation to learn to obtain the representation vector of the node.Secondly,an attention mechanism-based measurement node score prediction model is constructed,and finally the node representation vector is used as the input of the node score prediction model to predict the node score of the candidate nodes.This thesis comprehensively considers the node load,node importance,the impact of measurement on network performance,and the hardware and software consumption caused by measurement,and proposes a measurement node scoring standard that meets the network measurement requirements.In order to verify the effect of the algorithm in practical applications,this thesis builds a software-defined network based on a variety of commonly used network topologies,uses a random traffic model to generate link traffic,sets bandwidth and delay for the link,and simulates large-scale heterogeneity in reality.network and conduct experiments on this basis.At the same time,this thesis extends the proposed heterogeneous network node selection algorithm to the cross-domain environment,and combines it with federated learning,so that each autonomous domain can collaboratively train the model without exposing its own data.Experiments show that the heterogeneous network node selection algorithm can be applied to cross-domain and heterogeneous network environment.Based on the above research,this thesis implements a probe intelligent selection system,which is oriented to network measurement technicians and provides functions such as node selection,network management,model management,and user center.The system uses the Vue+SpringCloud framework to separate the front and back ends,follows the software development steps,and provides detailed requirements and design documents.Provide network measurement personnel with a user-friendly interface,convenient and quick operation visualization platform.