弹性光网络中组播相关的虚拟网络功能部署问题的研究

发布时间：2018-02-21 05:59

本文关键词： 弹性光网络跨数据中心网络网络功能虚拟化组播　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：众所周知,通过在跨数据中心的网络中运用网络功能虚拟化技术(NFV),服务提供商能够以更灵活有效的方式充分利用网络资源,并能缩短新服务的部署周期。本论文将研究如何在跨数据中心的弹性光网络中采用具有成本效益的策略来部署基于组播传输方式的树状NFV服务拓扑(简称组播NFV树)。本文试图联合优化虚拟网络功能(VNF)的部署与组播传输路径的路由与频谱分配,从而降低组播NFV树的服务开销。首先,本文研究了跨数据中心的弹性光网络中组播会话的动态构建问题。由于组播成员的动态变化会导致组播树性能的降低,本文提出了两种组播树选取策略,分别是DTS与QTS算法,用于选取出性能较低的组播树。针对选出的组播树,本文提出了两种组播树的调整算法,分别是完全组播树调整与局部组播树调整。实验结果显示,提出的算法能够有效的改善组播树的结构,降低网络阻塞率。同时,本文将算法在软件定义网络(SDN)平台上进行了验证。本文扩展了OpenFlow协议,使得网络能支持动态组播会话,并设计了控制平面的功能模块完成组播会话的动态调整。其次,本文研究了在跨数据中心的弹性光网络中组播NFV树的部署问题。本文考虑了静态网络规划与动态网络配置两种情况。对于静态网络规划,本文建立了一个混合整数线性规划(MILP)模型对问题进行求解,接着提出了三种启发式算法,分别是基于频谱槽辅助矩阵的算法AFM-MILP与AFM-GS,以及基于最短路径汇聚的启发式算法RB。仿真结果显示AFM-MILP与AFM-GS两种算法在服务开销上最为接近MILP算法,且运行时间远小于MILP算法。对于动态网络配置,本文考虑了频谱资源碎片,在AFM-GS与RB两种算法的基础上进行改进,设计了四种在线算法服务动态的组播NFV树请求。仿真结果显示,考虑网络中的频谱资源碎片能有效降低服务的阻塞率。当路由策略相同时,基于AFM-GS的算法性能优于RB算法。
[Abstract]:It is well known that by using network functionality virtualization technology in networks across data centers, service providers can leverage network resources in a more flexible and efficient manner. In this paper, we will study how to deploy the tree NFV service topology (multicast NFV) based on multicast transmission in flexible optical networks across data centers. This paper attempts to jointly optimize the deployment of VNFs and the routing and spectrum allocation of multicast transmission paths. In order to reduce the service overhead of multicast NFV trees, this paper first studies the dynamic construction of multicast sessions in elastic optical networks across data centers. Because of the dynamic changes of multicast members, the performance of multicast trees will be reduced. In this paper, two multicast tree selection strategies, DTS and QTS algorithms, are proposed to select the lower performance multicast trees. For the selected multicast trees, this paper proposes two multicast tree adjustment algorithms. The experimental results show that the proposed algorithm can effectively improve the structure of the multicast tree and reduce the network blocking rate. In this paper, the algorithm is validated on the software defined network (SDN) platform. In this paper, the OpenFlow protocol is extended to enable the network to support dynamic multicast sessions, and the functional modules of the control plane are designed to complete the dynamic adjustment of the multicast sessions. In this paper, we study the deployment of multicast NFV trees in elastic optical networks across data centers. In this paper, we consider static network planning and dynamic network configuration. In this paper, a mixed integer linear programming (MILP) model is established to solve the problem, and then three heuristic algorithms are proposed. The algorithms AFM-MILP and AFM-GSbased on the auxiliary matrix of spectrum slot, and the heuristic algorithm based on the shortest path convergence, respectively. The simulation results show that AFM-MILP and AFM-GS are the most close to the MILP algorithm in the service cost. The running time is much less than that of MILP algorithm. For dynamic network configuration, this paper considers the spectrum resource fragment, and improves it on the basis of AFM-GS and RB algorithms. The dynamic multicast NFV tree requests of four online algorithms are designed. The simulation results show that considering the spectrum resource fragments in the network can effectively reduce the blocking rate of the service. When the routing strategy is the same, the performance of the algorithm based on AFM-GS is better than that of the RB algorithm.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.1

【相似文献】