光数据中心网络中针对NFV服务链部署的多维资源联合优化研究

发布时间：2018-03-12 14:27

本文选题：弹性光网络　切入点：数据中心　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】：网络功能虚拟化(Network Function Virtualization,NFV)技术旨在利用数据中心里的通用设备实现网络功能,来替换传统的专有设备,从而实现网络功能的灵活部署和拆除,提升网络管理的智能化。此时的数据中心需要根据虚拟网络功能(Virtual Network Function,VNF)的部署情况,对内部资源进行调度,以提高资源利用率。对于弹性光数据中心网络(Inter-Data Center Elastic Optical Network,Inter-DC EON)中进行NFV服务链部署时,其数据流量大小在经过网络功能时通常可能变化。这种数据流量的动态变化,以及弹性光数据中心网络中的数据流量通常具有高吞吐量和高峰值且高突发性的特点,对底层物理结构提出了新的挑战。弹性光网络所提供频谱灵活的分配以及大容量的带宽,给弹性光数据中心网络中NFV服务链的部署提供了灵活可靠的底层支撑,而且,光网络中的NFV服务链还有低功耗等优点。但是,这种灵活性就要求上层对网络频谱资源和数据中心资源进行高效地管理和调度。本论文中,我们将首先讨论在Inter-DC EON网络中部署NFV服务链基本单元时出现的多维资源碎片,研究对频谱资源碎片和IT资源碎片的联合去碎片问题。为了提高Inter-DC EON网络中的资源利用率,我们联合考虑网络中的多维资源和网络重新服务的复杂性,对多维资源的分配进行重新优化。对于每一次联合去碎片,我们首先研究请求选择策略,并根据网络资源使用状态,提出自适应请求选择策略,即,自适应从频谱资源方面和IT资源方面选取请求。然后,我们对选出来的请求进行重新服务,首先利用混合整型线性规划模型进行数学建模,接着提出多种高效算法来解决数学模型的高复杂度。仿真结果表明,本论文中提出的联合去碎片算法可以有效地提高频谱资源和IT资源的利用率,从而降低Inter-DC EON网络中请求的阻塞率。接着,本论文详细研究了在弹性光数据中心网络中部署NFV服务链时,频谱资源与IT资源的联合分配问题。对于这个问题,我们将首先对问题进行数学建模,然后提出包括基于最长公共子序列的算法等多种高效算法来解决NFV服务链部署问题。最后通过仿真平台,评估提出的算法性能,验证了联合部署算法能够同时降低网络中的最大带宽槽下标和网络中部署的VNF数量,达到资源联合优化的效果。
[Abstract]:Network function virtualization Network Function Virtualization (NFV) technology aims to implement network functions by using universal devices in data centers to replace traditional proprietary devices, so as to achieve flexible deployment and dismantling of network functions. To improve the intelligence of network management, the data center needs to schedule the internal resources according to the deployment of Virtual Network function VNFs. In order to improve resource utilization. When deploying the NFV service chain in the flexible optical data center network Inter-data Center Elastic Optical Network, the size of the data traffic may change when it passes through the network function. As well as the characteristics of high throughput, high peak value and high burst of data traffic in elastic optical data center network, it presents a new challenge to the underlying physical structure. The flexible allocation of spectrum and large capacity bandwidth are provided by elastic optical network. It provides flexible and reliable underlying support for the deployment of NFV service chain in elastic optical data center network. Moreover, the NFV service chain in optical network has the advantages of low power consumption and so on. This flexibility requires the upper layer to efficiently manage and schedule the network spectrum resources and data center resources. In this paper, we will first discuss the multi-dimensional resource fragments that occur when deploying the basic unit of the NFV service chain in the Inter-DC EON network. In order to improve the resource utilization in Inter-DC EON network, we consider the complexity of multi-dimensional resource and network re-service in Inter-DC EON network, in order to improve the resource utilization in Inter-DC EON network, we study the problem of joint defragmentation of spectrum resource and IT resource fragment. For each joint defragmentation, we first study the request selection strategy, and propose an adaptive request selection strategy according to the network resource usage state. Then, we re-service the selected requests, and we first use the mixed integer linear programming model to build mathematical model. Then several efficient algorithms are proposed to solve the high complexity of the mathematical model. Simulation results show that the proposed joint defragmentation algorithm can effectively improve the utilization of spectrum resources and IT resources. In order to reduce the blocking rate of requests in Inter-DC EON networks, this paper studies in detail the joint allocation of spectrum resources and IT resources when deploying NFV service chains in flexible optical data center networks. We will first model the problem, then propose some efficient algorithms, such as algorithms based on longest common subsequences, to solve the problem of NFV service chain deployment. Finally, we evaluate the performance of the proposed algorithm through simulation platform. It is verified that the joint deployment algorithm can reduce the maximum bandwidth slot subscript in the network and the number of VNF deployed in the network at the same time and achieve the effect of joint resource optimization.
【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】：TP308;TN929.1

【相似文献】