Research On Virtual Network Resource Management Architecture And Mapping Methods For Future Networks

With the development of computational theories and technologies,new types of network are emerging.From the perspective of function and morphology,networks have changed from"Network Centralized" to "Application Centralized".Everything as a Service is becoming the ultimate goal of future networks.Due to lack of adaptability and resilience,the current Internet architecture can not meet the service diversification demand of users.Network virtualization is a key technology for the clean slate future networks design.By sharing a same substrate network,network virtualization technology can build multiple isolated and heterogeneous virtual networks,in which researchers can develop and test new network architectures and protocols.The node and link resources in each virtual network can be reasonable allocated according to the users' dynamic demands.Network virtualization can innovative the service mode and meet the service diversification demand of the users.Due to the idea of using network virtualization technology to build future network architecture proposed in a relatively short time,the corrsponding research is still in its infancy.How to virtualize the distributed,heterogeneous,autonomy,dynamic substrate network resources into virtual resources for unified deployment and management?How to build virtual networks through mapping method to meet the user service demands?These two issues are urgently to be solved for network virtualization.Accordingly,this paper aims at solving resource management architecture building and virtual network mapping problems for the future network.Based on the theory of cognitive computing,swarm intelligence,topological theory,social network centrality,this paper builds a hierarchical cognitive resources management framework,and proposes three new network virtualization mapping methods from the efficient,reliable and energy-saving the point of view,which provides references for the promotion and improvement of network virtualization technology.The main researchcontents of this paper are as follows.First of all,according to the construction problem of the resource management architecture in the network virtualization environment,a hierarchical cognitive resources management framework(HCRMF)is bulit,which has the characteristics of "centralized resource management,distributed task scheduling".The HCRMF has good scalability,flexibility,autonomy and reliability,which can reduce the difficulty and cost of management.There are different administrative domains bwetween layers which use the centralized management method.Cognitive units are embedded in task scheduling servers which can mutually cooperate and assign the task.This will reduce the human intervention,and make an efficient response to the users' virtual network request,and meet the users' individual service.The simulation results show that the HCRMF architecture can significantly reduce the pressure of the management nodes and links to enhance the virtual network requests' response speed compared with the existing centralized and distributed management architectures.Secondly,in order to solve the problems that existing network virtualization mapping algorithms which suffers from low efficiency and limited solution space,a virtual network embedding algorithm based on artificial fish swarm(VNE-AFS)was proposed under the premise that the substrate network resources are limited and the path splitting is not supported.A binary combinatorial optimization model was built according to the constraints on nodes and links between virtual network and substrate network,and the artificial fish swarm algorithm was used to achieve the approximate optimal mapping.The simulation results indicate that the costs of substrate network and computation time are reduced,and the success rate and average revenue of embedding are increased compared with the existing virtual network embedding algorithms.Thirdly,in order to solve the problem that virtual network is not reliable due to the substrate node failure,a reliable heuristic mapping method based on survivable virtual network(RHM-SVN)is proposed.We take the solution of the survivable virtual topology routing problem in dense wavelength division multiplexing optical network for reference to ensure that the rest of the mapped virtual networks keep connected under a single node failure condition in the substrate network,which guarantees the completeness of the virtual network and continuity of services.In order to reduce the cost of the substrate network,we formulate the reliable mapping problem as an integer linear program and the RHM-SVN is used to solve this problem.The simulation results show that the acceptance rate of virtual networks,the average revenue of mapping,and the recovery rate of virtual networks are increased compared with the existing reliable mapping algorithms,respectively.Finally,the current networks are designed for peak loads leading to low utilization of power resources.In order to slove this problem,a heuristic energy-saving virtual network embedding algorithm based on Katz centrality(Katz-VNE)is proposed.For sloving energy-saving virtual network embedding problem,the paper introduces Katz centrality to represent the node influence.In order to minimize the energy consumption of the substrate network,the energy-saving virtual network embedding problem is formulated as an integer linear program,and Katz-VNE is used to solve this problem.Katz-VNE tries to map the virtual nodes on the substrate nodes with high Katz centrality,which has been working,and uses the shortest paths offering the best bandwidths factor to avoid the hot nodes.Simulation results demonstrate that the long-term average energy consumption of substrate network are reduced significantly,and the long-term Revenue/Cost ratio,the acceptance rate of virtual network requests,and the hibernation rate of substrate nodes and links are improved significantly.