Performance Comprehensive Evaluation Based On Graph Entropy Of Virtual Network Embedding Algorithm

In the era of Internet of Everything,with the surge in network access,traditional network services are lacking in mobility,security,reliability and utilization efficiency.Network virtualization technology is a mechanism for sharing network resources derived from effectively solving the above problems.It allows multiple heterogeneous virtual networks to coexist on the same substrate network.In order to efficiently configure physical network resources,researchers have proposed a series of virtual network embedding(VNE)algorithms.With the emergence of a large number of VNE algorithms,the comprehensive evaluation of the performance of the VNE algorithm has also followed.There are three difficulties in the comprehensive evaluation of the performance of the VNE algorithm: firstly,how to determine the evaluation index system;secondly,how to analyze the relationship between the evaluation indicators;finally,how to construct the evaluation index information function.Analyzing the research dynamics of VNE algorithm,we can find that the research scholars mostly focus the research work on the direction of performance optimization of VNE algorithm,but compare the optimization points of the algorithm with other algorithms at the end of the research work.By reading a large number of references,this paper proposes to establish a VNE algorithm evaluation index system from ten aspects such as VNE cost,VNE revenue,VNE node pressure,VNE embedded link pressure,etc.From these ten aspects,we can comprehensively evaluate and analyze the research contribution of a VNE algorithm to virtual network embedding.For the problem established by the relationship between the evaluation indicators,since the relationship between the indicators is mutual,the matrix constructed by the relationship of indicators should be a symmetric matrix.In this paper it is proposed that the covariance matrix method can not only obtain the correlation coefficient between the evaluation indexes,but also reduce the high-dimensional data and obtain a symmetrical square matrix.Theadjacency matrix of the evaluation index can be obtained from the covariance relationship between the evaluation indexes,and the information function of each evaluation index can be constructed by the relevant graph entropy theory.In order to compare the performance of the VNE algorithm,it is necessary to obtain the comprehensive evaluation index data of the VNE algorithm performance.Based on the Alevin platform,four VNE algorithms are adopted to perform static virtual network embedding experiments,circularly embedded 20 times.And 20 sets of17 evaluation index data generated in the virtual network embedding process are obtained.Then,the original data is processed by the method described above,and the information entropy and weight of each evaluation index can be solved.Finally,constructing VNE algorithm Performance comprehensive evaluation function f,the better evaluation results show that the VNE algorithm improves the virtual network revenue and the virtual network request acceptance rate under the premise of reducing the physical network load,providing a direction for the optimization of VNE algorithm.