【作者】 李莉；

【导师】 魏翼飞；

【作者基本信息】 北京邮电大学 ， 电子科学与技术， 2020， 博士

【摘要】 随着移动网络流量的急剧增加,移动通信网络中的能耗问题日益严峻。出于经济利益和环境保护的双重考虑,提高网络中的能源效率成为第五代移动通信技术中设计无线通信系统的重要指标之一。广泛的研究表明,提高网络能效的有效措施包括:多类型接入设备组合覆盖蜂窝小区、可再生能源和传统能源协作供能和对不同类型的网络业务按需分配网络资源。因此需要对存在设备、能源和业务多样性的无线异构网络的特性进行研究。但是,由于无线异构网络采取垂直建站和分布式管理模式,很难根据全局网络状态对网络资源实现实时的优化分配。软件定义网络(Software Defined Network,SDN)架构通过分离网络的控制平面和数据转发平面对通信网络技术进行了改革,实现了灵活高效的可编程和集中化网络管理。网络功能虚拟化(Network Function Virtualization,NFV)作为软件定义网络的重要应用,通过分离网络功能和网络设备,实现了针对不同业务的虚拟网络划分机制。本论文针对无线异构网络的动态特性,以及目前因为管理困难造成的能效低下问题,利用软件定义网络技术提供的架构优势,实现了无线异构网络中灵活实时的网络资源调度配置,本课题的主要研究思路为,首先以网络性能和用户需求为约束,然后在此基础上降低网络能耗,以此提高网络能效。具体研究工作与贡献如下:1.系统架构分析通过对软件定义网络架构的分析,针对无线网络中的设备异构性和能源异构性,提出一个基于SDN技术的绿色中继网络架构的设计方案,为动态的无线网络提供灵活的网络管理模式;针对无线网络中的资源异构性和业务异构性,提出一个基于NFV技术的智能虚拟边缘网络架构的设计方案,智能的实现了网络资源的按需分配。该架构中的虚拟网络运营商将网络业务提供商提供的实体资源整合后,以虚拟网的形式按照需求提供给网络服务供应商,并允许对边缘网络资源进行智能调度。2.能效优化的流表项管理策略在绿色中继无线网络中,用户本应按照最节能的方式选择基站或者中继节点接入网络,但是可再生能源捕获量难以预测,当中继节点中的可再生能源出现短缺时,该设备无法为用户提供接入服务,在这种情况下,接入该节点的用户需要切换到基站上。另一方面,当可再生能源的捕获量超过消耗量,多余的能源将由于无法被储存而浪费。针对可再生能源的中断和饱和问题,本文提出一种软件定义网络架构下的动态流表项更新策略和一个面向长期能效的流表管理策略,根据节点的可再生能源状态为用户选择接入点,以最大限度的利用已经捕获的可再生能源,从而减少传统能源的消耗量,实现传统能源的能效优化。这些策略可以集成为软件定义网络的控制器上的一个网络管理应用程序。数值仿真结果表明和其他路由策略(例如容量受限和最优路径策略)相比,所提出的策略可以有效利用可再生能源,提高传统能源的能效。3.能效最优的Ant-Q缓存数据流分发策略在移动边缘网络部署缓存功能可以显著减少边缘网络和远端云之间的通信,从而节省能源,然而由于用户的移动性、业务多样性和有限的边缘存储计算能力,如何智能的使用边缘节点的通信和缓存功能具有很大的挑战性。本文使用Ant-Q学习算法为虚拟化移动边缘网络提出了一种节能感知的动态缓存策略。针对不同业务的需求差异,本文首先设计一种具有边缘缓存功能的虚拟化移动边缘网络范例,和直接从移动用户行为中提取特征相比,本范例可以更准确有效地从网络服务中提取特征,并建立其系统模型。然后将能效感知的缓存数据流分发问题根据强化学习算法建模,并使用结合强化学习和蚁群算法的Ant-Q算法解决该问题,该算法可以有效结合云计算能力和边缘网络的高响应能力。此外本文使用禁忌空间来提高Q学习算法的收敛性。仿真结果表明,所提出的缓存数据流分发策略可以显著提高能源效率,同时对网络性能进行优化,而带禁忌空间的Q-learning算法可以显着减少计算时间,提高Ant-Q学习阶段的计算效率。4.能耗最低的边缘虚拟网络构建为了支持具有高资源利用率的异构无线网络中的多个设备和服务的无缝通信,网络虚拟化技术提供了灵活和可扩展的管理。本文研究了无线多跳边缘网络中虚拟网络构建问题,首先分析无线多跳蜂窝网络场景,并建立虚拟网络构建模型。之后将该问题转化为多商品流问题,提出了一种基于多商品流算法的最小成本流算法,即用最小的能耗开销相应虚拟网请求,最后采取拉格朗日松弛优化算法和次梯度算法来解决该问题。仿真结果表明,所提出的多商品流算法可以在无线多跳网络中带宽和节点发射功率有限的情况下提高虚拟网络请求的接受率,以此提高网络能效。更多还原

【Abstract】 With the rapid exploding of mobile internet traffic,the problem of energy consumption in mobile networks is becoming increasingly serious.Because of energy cost and environ-mental concerns,energy efficiency have become key performance metric for designing wire-less communication systems of the fifth generation mobile communication system(5G).Ex-tensive research shows that effective measures to improve network energy efficiency include:multiple types of access equipment combination covering cells,cooperative energy supply of renewable energy and traditional energy and allocating network resources on demand for different types of network services.Therefore,the characteristics of wireless heteroge-neous networks with diverse equipment,energy,and services need to be studied.However,due to the wireless heterogeneous network adopts vertical station establishment and dis-tributed management mode,it is difficult to achieve real-time optimal allocation of network resources according to the global network status.Software Defined Network(SDN)archi-tecture reforms the communication network technology by separating the control plane and data forwarding plane of the network,and realizes flexible and efficient programmable and centralized network management.As an important application of software-defined network-ing,Network Function Virtualization(NFV)technology realizes a virtual network division mechanism for different services by separating network functions and network devices.In order to solve the problem of low energy efficiency caused by management difficulties in wireless heterogeneous networks,this topic uses the architectural advantages provided by software-defined network technology to reduce network energy consumption on the premise of ensuring network performance by dynamically scheduling network resources.Specific research work and contributions are as follows:1.System architecture analysisBased on the architecture of software-defined networks,aiming at device heterogeneity and energy heterogeneity in wireless networks,a design idea of software-defined green relay network architecture based on SDN technology is proposed.The architecture is able to offer flexible management for dynamic wireless network environments.Aiming at device hetero-geneity and service heterogeneity in wireless networks,a virtual edge network architecture based on NFV technology is proposed.The virtual network operator in this architecture integrates the physical resources provided by the network service provider and provides it to the network service provider in the form of a virtual network according to demand,and allows intelligent allocation to network resources.2.Energy efficient-oriented flow-table management strategyIn the beginning,all the users can connect to a BS or an RN in the most energy efficient way,however,the availability of renewable energy is unpredictable,and users connected to an RN will have to be handed over to BS when there is not enough renewable energy to power the RN.There are also occasions when an RN harvested more renewable energy than required to server the users connected to it,in these cases users connected to BS or other RNs can be transferred to connect through this RN,otherwise the harvested surplus renewable energy will be wasted as it cannot be stored.We transform the problem of optimizing the use of harvested renewable energy into a flow-table management problem.A model of harvested renewable energy available for RNs is built first,then we propose a dynamic flow-table updating strategy and a long-term energy efficiency-oriented flow-table management strategy,which are driven by the availability of renewable energy,to maximize the use of the harvested renewable energy and thus minimize the use of traditional energy.The proposed strategies can be integrated as applications on top of an SDN controller.Simulation results show that the proposed strategies can make more efficient use of renewable energy thus make the overall relay network more energy efficient compared with other routing strategies such as the capacity limited strategy(CLS)and the optimal path strategy(OPS).3.Energy efficient cached flow distribution policy for edge networks using Ant-Q learningDeploying edge caching ability can significantly reduce the traffic volumes between edge networks and cloud,so as to save energy.However,how to intelligently use the com-municating and caching capabilities of edge nodes/cloudlets is challenging because of users’mobility and limited edge compute power.This paper proposes an energy aware dynamic cache policy for virtualized mobile edge networks using Ant-Q learning algorithms.We first design a paradigm of virtualized mobile edge networks with edge caching ability,which can more accurately and efficiently extract characteristics from network services than from mo-bile users’ behavior,and build its system model.Then we formulate the energy-aware cache optimization problem into a reinforcement learning(RL)model,and solve the problem with a dynamic cache policy use Ant-Q algorithm,which can efficiently combine strong comput-ing capacity of cloud and the responsiveness of edge networks.Furthermore,we use tabu space(TS)to improve the convergence of Q-learning algorithm.Simulation results show that the proposed dynamic cache policy can enhance energy efficiency and improve network performance,besides,the proposed Q-learning algorithm with TS can significantly decrease computation time and improve computation efficiency in learning stage.4.Edge virtual network embedding with the lowest energy consumptionTo support seamless communication of multiple devices and services in heterogenous wireless networks with high resources utilization,network virtualization has been proposed to offer a flexible and scalable management.In this paper,we study the wireless multi-hop edge network and the problem of virtual network embedding.We first analyze the wireless multi-hop cellular network scenario and establish a virtual network embedding model.After that we propose a minimum cost flow algorithm based on the multi-commodity flow algo-rithm.Thus the problem is transformed into a multi-commodity flow problem.We finally put forward an optimization algorithm of Lagrange relaxation and sub-gradient algorithm to solve the problem.The simulation results show that the proposed multi-commodity flow algorithm can make full use of network resources,and improve the acceptance rate of virtual network requests with limited network resources,so as to improve the energy efficiency.更多还原