Research On The Key Technology Of Dynamic Resource Adaptation Based On Wireless Network Virtualization

With the rapid development of wireless communication technology,wireless networks are gradually penetrating into people’s daily lives,and wireless service types and application scenarios have become diversified.Considering maneuver communication scenarios,such as medical rescue and military communication,etc.,wireless networks should have the ability to be rapidly deployed and to provide on-demand wireless services for users.However,under the traditional rigid network architecture,the network and services are coupled together,making it difficult to adapt to the differentiated requirements of users.Based on wireless network virtualization,network resources can be decoupled from the underlying physical devices.In addition,the resource utilization and flexibility of the wireless network can be improved by dynamic resource adaptation.Three major challenges in resource adaptation based on wireless network virtualization are summarized as follows.(1)Multidimensional:Simultaneous deployment of heterogeneous networks brings multiple types of resources.The abstraction of network resources and the on-demand construction of virtual networks need to be addressed first.(2)Highly dynamic:In maneuver communication scenarios,the high mobility of nodes,environmental changes,and unstable wireless links all lead to a highly dynamic wireless environment.Network resource adaptation algorithms are required to have stronger environmental adaptability.(3)Differentiation:The quality of service(QoS)requirements for diverse wireless services are highly differentiated.Therefore,the accuracy and applicability of network resource adaptation algorithms need to be improved.To address the above challenges,this dissertation first proposes an abstract representation model of network resources and an on-demand construction method for virtual networks.Then the network resource dynamic adaptation technology is further studied,from both carrier network and access network perspectives.The main contributions and innovations are shown as follows.(1)Research on the on-demand construction technologies of virtual networksIn order to address the issues of difficult multidimensional network resource management and rigid virtual network construction methods in maneuver communication scenarios,a model for abstraction and representation of network resources is first proposed.Specifically,the control interface for different kinds of services is provided for fast identification by designing the maximum flow-based and maximum value-based resource abstraction methods.The proposed approach achieves the abstraction of complex physical resources and presents a unified service capability for users.Then,dimensionality reduction of network resources can be achieved by building local virtual resource pools and global virtual resource pools.The overhead of synchronizing network resources can be reduced significantly.Based on this model,an on-demand virtual network construction method is proposed to achieve the adaptation of network requirements and virtual resources.Specifically,a topology optimization algorithm based on the betweenness of virtual links is proposed,which reduces the overhead of virtual network construction.Then,a hierarchical resource reservation scheme is proposed to enhance the service capability of the virtual network for service parallel transmission scenarios.Simulation results indicate that the proposed method can reduce the construction overhead by 38%and effectively improve the network resource utilization while ensuring efficiency.(2)Research on resource adaptation technologies for carrier networks based on wireless network virtualizationConsidering the higher requirements of resource adaptation capability in a highly dynamic environment for maneuver communication scenarios,resource adaptation algorithms for carrier networks are studied in this dissertation.The cluster head centralized and node exploration distributed path resource adaptation algorithms are designed for complete and missing global resource information scenarios,respectively.These two designed algorithms achieve efficient on-demand matching between differentiated requirements and network resources in dynamic environments.For the scenario with complete global information,a cluster head cooperative routing algorithm based on deep reinforcement learning is first proposed,to meet the time-varying wireless channel and differentiated requirements of services.The proposed algorithm learns the optimal routing strategy in the current network state by off-line training.By learning channel characteristics and user requirements and optimizing the training process through historical observation information,the proposed algorithm can effectively avoid network congestion and further improve the adaptability of the routing algorithm in a highly dynamic environment.Then,a routing algorithm based on distributed RL is proposed for the scenario with missing global information.By the means of information iteration between member nodes,the distributed routing decision is achieved.Basic communication service is guaranteed in the extreme condition of cluster head failure.The simulation results show that the proposed resource adaptation algorithm for the carrier network can effectively reduce the end-to-end delay of services.In addition,the proposed algorithm can improve the utilization of network resources and the capacity of the network by avoiding congestion.(3)Research on resource adaptation technologies for access networks based on wireless network virtualizationAiming at multi-user differentiated QoS requirements for access networks and the problem that it is difficult to accurately match multidimensional network resources,a node selection algorithm based on distributed reinforcement learning for access networks is proposed to improve network resource utilization first,which realizes the node selection based on user decision and the combination optimization of computing and wireless resources.Specifically,the multi-objective problem of content delivery delay and system power consumption performance is established,to ensure the differentiated QoS requirements of different virtual networks.Simulation results show that the proposed method can dramatically reduce the content delivery delay and improve the utilization of network resources.Afterwards,in order to provide finer-grained network resource on-demand matching for different types of users,a joint allocation algorithm of wireless subchannel and node transmit power is further proposed.A multi-performance weighted utility function is defined to describe the differentiated requirements of virtual networks.The dual decomposition method is used to solve the NP-hard mixed integer programming problem,and a suboptimization algorithm is proposed to significantly reduce the computational complexity of network resource adaptation.Simulation results show that the proposed joint allocation method can improve the on-demand service capability of access networks and improve the utilization of network resources while ensuring the requirements of different virtual networks.