Study On Complex Service-Oriented Key Network Virtualization Techniques

In recent years,complex network applications and business requirements with new features such as diversification and differentiation are emerged,with the rapid development of the Internet of Things and artificial intelligence.In addition,the capabilities and efficiencies of existing networks are difficult to fulfill the new demands.How to solve the contradiction between the networks of various systems with limited capabilities and efficiency and the complex and diverse business needs is one of the key issues in the field of communication networks.In recent years,network virtualization has become an effective approach to solve the above problems.Network virtualization has received extensive attention from both academia and industry.Network virtualization enables multiple virtual networks(VNs)to co-exist on a physical network(also known as the underlying network,SN),according to VNs' application and service requirements.The virtual network can run with various customized network protocols.Network virtualization can implement flexible and diversified network services without changing the underlying network architecture.At the same time,network resource utilization can be greatly improved by sharing multiple underlying networks.Network virtualization has become one of the key attributes of the future network.In network virtualization,multiple virtual networks share resources such as nodes and links of the underlying network.Each virtual network has its specific resource requirements such as computing,storage,and communication bandwidth.Therefore,under the premise of meeting different virtual network resource requirements,how to embed the virtual network onto the substrate network efficiently has become one of the key issues for network virtualization.Mapping virtual network requests onto the underlying network is called as Virtual Network Embedding(VNE).VNE determines the resource utilization of the underlying physical network and has important potential significance.The VNE solution is proved to be an NP-hard problem and cannot be solved in polynomial time.Therefore,VNE is one of the core technologies for network virtualization,and it is also a research hotspot in network virtualization research.This thesis conducts a comprehensive research on virtual network embedding(VNE)algorithms.Three novel VNE algorithms are proposed based on nodes'(substrate and virtual)importance,linear programming model,and candidate sets aid.Main ideas of three VNE algorithms are shown below:1.A novel VNE algorithm,based on the nodes' importance,is proposed.The aim of proposed algorithm is to compare with previous algorithms which only consider single topology and local node resources.The proposed algorithm is a new collaborative two-step(node-link)embedding algorithm.Compared with previous algorithms that merely consider the node's adjacent link bandwidth,our algorithm considers the nodes' importance and link propagation delay while embedding VNs.Iterative approaches are assisted to measure the(physical and virtual)node importance before embedding each given VN.In addition,resource attributes and topological attributes of each node are quantified so as to improve the VN embedding efficiency.The link propagation delay attribute further improves the quality of VN embedding.The algorithm based on iterative approaches can calculate and rank the importance of each node,and then uses greedy algorithm and the shortest path approach to complete the node embedding and link embedding.The node capacity,the link bandwidth constraints and link propagation delay constraints are considered while conducting the VN embedding.Simulation results show that the proposed algorithm can reduce the average link propagation delay while ensuring VN acceptance rate and revenue to cost ratio.2.With the goal of achieving the optimal embedding solution per given VN,novel VNE algorithm(VNE-ILP),based on linear programming model approach,is proposed.Different from previous heuristic algorithms,it constructs an integer linear programming model to solve the VNE problem.Considering the link propagation delay attribute and node location attribute,the algorithm constructs a network model based on the mathematical model based on ILP,with constructing reasonable objective functions and constraints and giving a reasonable convergence accuracy and threshold.Finally,VNE-ILP obtains the optimal or local optimal virtual network embedding results.Compared with previous heuristic algorithms,our proposed algorithm VNE-ILP not only considers minimizing physical resources consumption and achieving load balancing,but also minimizing the average delay with guaranteeing the receiving rate.In addition,node location is considered.Simulation results vividly show that VNE-ILP performs well in small and medium-sized networks,in terms of receiving rate and cost-benefit ratio.3.We propose another novel VNE algorithm based on candidate node path aid,aiming at removing the high complexity of pure ILP based programming.Its core idea is to construct the candidate node and link sets before conducting the integer linear programming(ILP)based embedding.According to the properties of the real network,our proposed algorithm considers four different node-link constraints.These node-link constraints are node capacity constraint,node location constraint,link bandwidth constraint and link propagation delay constraint.Compared with the existing heuristic algorithms,our proposed algorithm can guarantee that the quality of any embedded VN is optimal.In addition,the time complexity of our algorithm can be significantly reduced,compared with the pure ILP approach.It lies to the fact that we construct candidate substrate nodes and substrate path subsets.Simulation results demonstrate that our proposed algorithm outperforms selected typical heuristic algorithms.