【作者】 李琳；

【导师】 郑侃；

【作者基本信息】 北京邮电大学 ， 信息与通信工程， 2019， 博士

【摘要】 无线通信技术的快速发展给人们带来了丰富便捷的移动体验,改变了日常工作和生活方式。然而,为满足日益增长的业务需求和日新月异的技术变革,无线网络的部署逐渐趋于异构化和密集化,这种趋势一方面增加了移动运营商的资本和运营开销,另一方面对无线网络的管理和维护带来了新的挑战。无线网络虚拟化(Wireless Network Virtualization,WNV)是应对以上问题的重要手段之一。它打破了传统无线网络架构的僵化特征,以抽象、共享、分配和重构的方式实现按需、高效的资源管理来降低移动运营商的开销,以功能解耦和数据控制分离的方式实现异构设备的集中控制和灵活管理。本论文在WNV的研究中以面向动态、多样化的业务需求为方向,对终端和网络实现按需的联合控制和资源管理。主要工作内容概括如下:1.提出了一种基于网络虚拟化的联合管理架构。本文在架构设计中,除了包含空口分离、虚拟基站和软件定义无线接入网等传统虚拟化设计外,还综合考虑了不同终端设备的能力特点和多终端参与的业务形式(Multiple Endpoints Service,MES)。架构的设计思路和内容有:首先,采用基于软件定义网络的三层架构的设计,引入北向、南向接口实现中心控制与网络应用、底层基础设施的编排交互;其次,在网络管理层中引入了模块设计,包括网络管理、优化控制和服务管理,通过模块间的相互作用保障架构的灵活性和可行性;,最后,在优化控制模块中引入了终端控制保障、服务资源分配和网络负载均衡策略,实现无线网络的无缝、可靠的联合管理和高效、按需的资源分配。2.提出了一种面向控制面可靠性保障的动态资源划分方案。本文主要依据空口分离的技术方案,即:在宏、微小区资源正交的异构网络中,宏基站为终端提供控制面的信令传输,而微基站提供数据面的业务承载。对于正交频谱的划分问题,本文采用动态的策略,在终端控制面信令阻塞率小于目标的前提下,最大化微小区划分的频谱带宽来提升系统吞吐量性能。本文通过马尔科夫模型分析得出分配带宽与控制面阻塞率的计算表达式。根据不同场景下的业务量模型,仿真验证了提出的动态资源划分方案能够在保障终端控制面可靠性的前提下,同时提高系统的吞吐量。3.提出了一种基于虚拟网络映射(Virtual Network Embedding,VNE)的资源管理方案。该方案实现了本论文设计的联合管理架构中的联合优化策略:终端、接入网和核心网的联合按需分配。本文在联合优化问题上根据虚拟网络业务(Virtual Network Service,VNS)对网络功能需求和终端需求分别解耦为服务功能链VNE和终端流量图VNE问题。本文针对涉及到无线网络特征的VNE问题,通过以下两方面展开:(1)对无线资源、计算资源和回传带宽的联合优化VNE问题建模,应用了Leontief经济学理论,对多目标的VNE问题进行转换,形成混合整数线性规划问题求解;(2)对VNS多种服务质量(Quality of Service,QoS)需求条件下的系统收益最大化VNE问题建模,应用有效带宽理论中无线信道特征与数据传输QoS需求的关系模型,根据业务在不同QoS等级的收益情况以VNE的形式提供具有服务质量保障的带宽资源分配,此外针对不同业务的优先级策略,并提高算法的计算效率,设计了一种基于优先级策略的局部梯度检索算法。4.提出了基于虚拟网络迁移的资源均衡技术。在本论文设计的联合管理架构中,由于终端的移动性、无线链路的不稳定性,很难保障系统性能的高效可持续性,据此本文以虚拟网络迁移为手段来动态地均衡网络资源。本文综合考虑了网络中无线资源池的特征:(1)无线链路间的隔离性较差(2)数据传输与资源消耗在数量上非一致性,将虚拟网络迁移建模为非凸优化问题,并采用了两种不同的近似转换方法分别求解,仿真结果验证了提出的两个迁移算法相对传统迁移算法具有更显著的均衡效果。更多还原

【Abstract】 With the rapid development of wireless communication technology,diversified services and expanded scale have made the deployment of the network denser and more heterogeneous.This leads to increased capital expenses and operation expenses of service operators.Wireless Network Virtualization(WNV)is a promising solution to this issue,because it can provide ways of flexibly control the physical hardware and fully utilize the network resource,by abstraction,sharing,allocation and re-construction of the resource,and by efficiently managing heterogeneous devices in ways of function decoupling and data control separation.This thesis investigates on~WNV targeting on variety and dynamic types of services to design efficient and effective schemes for on-demand joint control and resource management of UE and the network.The main contributions of this thesis are as follows:1.Proposed a VWNV-based integrated management infr-astructure.When designing the infrastructure,we not only take the traditional components into account including the decoupled air interface,virtual base station and software-defined radio access networks,but also different capabilities of different User Equipment(UE)and Multiple Endpoints Service(MES).Therefore,the proposed infrastructure has benefits for thr-ee key roles:for UE,it can decide the role that each UE takes in MES according to its capability and characteristics;for services development,service provider can focus on the design of the service rather than the distribution of UEs;for network service provider,the joint-optimization of both UE and the network can help a lot on dealing with dynamic and diversified service requirements and bring high r-evenue for them.2.Proposed a control plane reliability guaranteed dynamic resource allocation scheme.This thesis adopted the decoupled air interface technology to let macro base stations to provide UEs with continuous QoS-guaranteed control services wherever they are,in heterogeneous network where the resources of macro/pico cells are orthogonal to each other.Moreover,the allocation of the orthogonal frequency resource is done by a dynamic scheme:maximizing the available frequency resource of pico cells to guarantee the system throughput,under the constraint that the UE control signaling blocking probability should be less than the target probability in the coverage of macro base stations.We applied Markov decision process to derive the mathematical formula of the blocking probability.By applying the formula,we obtain the simulation results under different scenarios,and they demonstrate that the proposed resource allocation scheme can highly elevate the system throughput while guarantee the reliability of UE control plane in the mean time.3.Investigated Virtual Network Embedding(VNE)-based resource allocation scheme that implements the joint-optimization method in the proposed integrated management infr-astructure in the first contribution.This scheme adapts to the trend of more marginalization of network functionality and more diversified types of network services.Under this resource management architecture,all kinds of services are abstracted to Virtual Network Services(VNS).The network will provide services to them by the virtual network embedding of both Service Function Chain(SFC)and UE Traffic Graph(UT-Graph).To deal with the VNE challenges above,this thesis addresses the problem from two perspectives and solve them separately.First,we built the consumption models of various types of resources to fully optimize,and applied Leontief s principle to transform the original multi-objective VNE problem into mixed integer linear programming.Second,we incorporated multiple QoS of services into the system models to construct the VNE-based optimization problems under different scenarios,and maximized the system revenue by allocating resource for different services according to their own revenues while guarantee the QoS.Moreover,we proposed a Partial Staircase Search algorithm to prioritize different services and to effectively improve the computational efficiency of the algorithm.4.Proposed a VNS migration-based dynamic resource balancing scheme.Under the proposed integrated management infrastructure,the mobility of UE and instability of wireless links leads to discontinuous performa nce guarantee of the joint-optimization.To address this problem,we applied VNS migration to dynamically balance the network resource.We proposed an effective evaluation metric of VNS migration,and model the migration process as a non-convex optimization problem,taking into account the asymmetry of resource consumption and traffic flow in wireless networks.To solve the optimization problem,this thesis also proposed two different algorithms,and the simulation results illustrate that the proposed algorithms are much more effective on balancing the resource than the traditional heuristic algorithm of VN migration.更多还原