空分复用弹性光网络中串扰感知的虚拟网络映射算法
In recent years, the rapid increase in network traffic requires more network transmission capacity. Therefore, an elastic optical network based on orthogonal frequency division multiplexing technology is proposed. Its characteristic is to allocate spectrum resources according to the connection requests. By improving the utilization efficiency of spectrum resources in the frequency domain, the network transmission capacity can be expanded. However, restricted by the Shannon limit, the traditional single-core single-mode fiber is difficult to further expand the transmission capacity in the network. Therefore, on the basis of the proposed multi-core optical fiber (MCF), space division multiplexing elastic optical networks (SDM-EONs) are considered as one of the most potential solutions to break through the network capacity bottleneck. At the same time, the applications in the network have diversified demand characteristics, but it is difficult for physical facility providers to flexibly configure the network according to the specific needs of the applications. In recent years, the increasingly developed network virtualization technology has solved the problem of network rigidity and formed a diversified network structure. The combination of network virtualization technology and SDM-EONs can improve the utilization rate of the underlying physical resources, but it will make the virtual network mapping more complex. Furthermore, virtual network mapping inevitably needs to meet more constraints. There are also physical layer constraints of inter-core crosstalk (XT) in SDM-EONs. When the overlapping spectral segments of adjacent cores are occupied by different optical paths at the same time, the transmission quality of optical paths will be degraded, and high-quality transmission of normal signals cannot be realized. Therefore, this paper focuses on SDM-EONs based on MCF to solve the virtual network mapping problem of crosstalk perception.
In this paper, the crosstalk impact of the occupied spectrum resources on the adjacent core spectrum is considered, and a crosstalk impact assessment method is designed. Then a virtual network mapping algorithm for crosstalk awareness (CA-VNM) in SDM-EONs is proposed. The virtual network mapping method adopted by the algorithm is two-stage mapping. The first stage is virtual node mapping. In order to enhance the correlation between virtual nodes and virtual links, virtual nodes adopt the virtual node proximity ranking method, and physical nodes adopt the node degree classification method, so as to improve the virtual network acceptance rate and provide the underlying physical resources for the subsequent virtual services with greater demand. The priority evaluation method of physical nodes considers the XT on the adjacent links of nodes to improve the utilization rate of the underlying physical resources and balance the XT. The second stage is virtual link mapping, and the virtual link bandwidth is prioritized. Based on the prioritization, the K-shortest path method is used to find the K optical paths with the least hops, and the physical link priority method is used to rank the candidate physical links. Then the core load balancing and spectrum partition allocation methods are designed to balance the traffic of each core and reduce the XT.
The algorithm proposed in this paper optimizes the network performance. The request acceptance rate of CA-CVM achieves the best effect in the comparison algorithms, and the highest request acceptance rate of the virtual network is increased by 11% (Fig. 6). CA-CVM obtains the best spectrum utilization performance in the NSFNET network, and the spectrum utilization is increased by 17.1% (Fig. 7). The XT improvement rate gradually decreases with the increase in the load because the idle resources in the network gradually decrease, and other methods cannot reduce the XT. The XT improvement rate of CA-CVM is the best. At 100, 200, 300, 400, 500, 600, 700, 800, and 900 Erlang, the XT improvement rates are 67.7%, 54.6%, 46.3%, 39.3%, 32.8%, 20.5%, 11.8%, 9.1%, and 7.0%, respectively (Fig. 8).
In this paper, a virtual network mapping algorithm for crosstalk awareness in SDM-EONs is proposed, and the constraints of virtual network mapping in SDM-EONs are considered. When the priority of virtual nodes is evaluated, the proximity between virtual nodes is considered, and a virtual node proximity ranking method is proposed. When candidate physical nodes are selected, the physical node degree is considered so that more underlying physical resources can be provided to the virtual request service with greater demand. A crosstalk impact assessment method is proposed to evaluate the weight of physical nodes and balance the crosstalk between cores. In addition, the core is selected according to the bandwidth required by the virtual link to balance the core load. Finally, the spectrum is divided into several regions, and the first hit and the last hit are used alternately when the spectrum is allocated to reduce the influence of XT. The simulation results show that when the network load is small, the acceptance rate of the virtual network is optimized by 11%, the spectrum utilization rate is increased by 17.1%, and the XT improvement rate can reach 67.7%. In view of this, the proposed algorithm can effectively improve the acceptance rate of the virtual network and the XT in the SDM-EONs and optimize the mapping performance of the virtual network.
1 引言
近年来高清视频、云计算以及5G通信系统等新兴业务的蓬勃发展,网络流量急剧增加,这必然对网络传输容量有更高的要求。为此,研究人员提出基于正交频分复用技术的弹性光网络(EONs)。与传统的波分复用光网络相比,EONs根据连接请求的需求为其分配频谱资源,通过在频域提升频谱资源的利用效率,实现了网络传输容量的扩展[1-2]。然而,受到非线性香农容量极限的限制,传统单芯单模光纤难以进一步拓展网络的传输容量[3],基于多芯光纤(MCF)的空分复用弹性光网络(SDM-EONs)被提出。SDM-EONs被视为突破网络容量瓶颈问题的最具潜力的解决方案之一[4]。同时,网络应用具有多样化需求特征,但针对应用的特定需求,物理设施提供商很难灵活地配置网络。近年来网络虚拟化技术的快速发展,便于网络管理者抽象物理网络资源和灵活地部署网络应用,重点解决网络僵化问题,形成多元化的网络结构[5]。网络虚拟化技术与SDM-EONs相结合,可以提高底层物理资源利用率,但是会使虚拟网络映射问题更加复杂[6]。在SDM-EONs中,光信号在纤芯中传输时会有小部分信号功率泄漏至纤芯包层,影响相邻纤芯中占用相同频谱位置的信号,形成芯间串扰(XT),芯间串扰随着传输距离的增加而加剧,对信号的传输质量产生劣化影响,严重时会导致光路因不满足光纤传输质量的要求而无法正常传输。因此,基于SDM-EONs的虚拟网络映射需要满足的约束条件比底层网络是EONs的映射多一个串扰约束条件[7],而针对SDM-EONs中的芯间串扰,研究串扰感知的虚拟光网络映射问题具有重要的意义。
Xuan等[8]提出基于多芯纤芯SDM-EONs的虚拟网络映射算法,并通过改进全局优化算法来解决多芯光纤弹性光网络中的虚拟网络映射问题。然而芯间串扰是基于MCF的SDM-EONs中主要的物理层损伤来源,该算法没有考虑芯间串扰,未解决SDM-EONs中的关键问题。Chen等[9]引入了链路重要度和节点重要度的概念,定义了虚拟光网络(VON)映射优化的计算公式。为了降低频谱碎片化程度,他们引入了分段的概念,在虚拟链路映射之前,将虚拟链路的带宽需求划分为几个线速率,并在此基础上建立了3种启发式VON映射方法和2个整数线性规划(ILP)模型,然而该算法只考虑了串扰约束,未能减小SDM-EONs所产生的芯间串扰。刘焕淋等[10]在设计物理节点优先级公式时,考虑了物理节点的相邻链路上频谱资源的精确匹配度属性和候选物理节点到已被映射物理节点的跳数距离,从而降低了频谱资源浪费并且提高了虚拟网络映射成功率,但他们未重点考虑芯间串扰问题,只是采用纤芯优先级排序方法。Chen等[11]提出一种将最大资源需求的虚拟节点优先映射到资源最多的光节点的最短光路映射算法(LCLC),然而其底层光网络是弹性光网络。Liu等[12]提出频谱状态评估方法来评估当前网络中的芯间串扰,在频谱分配过程中,选择串扰较弱的频谱块进行分配,有利于后续业务请求持续分配。Zhang等[13]提出全网串扰辅助图来评估链路所受串扰的影响,实现了对全网串扰影响的动态评估,确保了全网串扰影响均衡。刘仕鑫[14]提出了频谱状态评估方法并对频谱进行区域划分,有效地减小了芯间串扰的影响。文献[12-14]提出的芯间串扰评估方法都没有体现出多芯光纤中最中间的纤芯对周围纤芯串扰的影响最大。在现有的相关工作中,研究者已经提出一些SDM-EONs与虚拟网络结合的算法[8-10],但是未见对虚拟光网络映射中的串扰问题以及芯间串扰评估的报道。
本文聚焦基于MCF的SDM-EONs解决串扰感知的虚拟网络映射问题。当芯间串扰超过串扰阈值时,不满足SDM-EONs中的串扰约束条件,导致虚拟网络映射失败。因此,本文提出一种基于SDM-EONs 的串扰感知虚拟光网络映射算法,通过减小芯间串扰提高虚拟光网络的请求接受率、频谱利用率和串扰改善率。
2 SDM-EONs下的虚拟网络映射模型
2.1 网络模型
底层SDM-EONs被抽象为无向图
2.2 约束条件
通常,虚拟网络映射问题需要遵循对应的约束条件。在虚拟节点映射过程中,除了物理节点的计算资源需要满足虚拟节点的计算资源外,还需要满足的约束条件分别为
式中:
在虚拟链路映射过程中,物理链路的可用带宽需要满足虚拟链路带宽需求;在频谱分配过程中,需要满足频谱一致性、连续性、不重叠的约束条件。此外,网络节点不进行纤芯转换,即在属于其路径的所有MCF链路中,将同一个纤芯分配给一条光路,这就是所谓的空间连续性约束。
式中:
除此之外,
式中:
为了确保所有分配到一个链路的串扰值小于其串扰阈值(
式中:
3 串扰影响评估方法
芯间串扰是SDM-EONs传输质量的最重要影响因素之一。芯间串扰是信号在纤芯间传输时一部分功率泄漏所产生的,且其在包层中传播时,泄漏的功率大小会随纤芯距离呈指数式下降。因此,本文仅考虑多芯光纤中的相邻纤芯占用相同频隙进行资源传输时会产生芯间串扰,而不相邻纤芯间的串扰忽略不计。由多芯光纤的纤芯构造可知,中心纤芯所产生的芯间串扰更大[14],为此对芯间串扰影响的评估需要增加中心纤芯所产生串扰影响的权重值。本文将多芯光纤中的光纤分为两类,第一类为中间光纤周围的纤芯
本文采用七芯光纤SDM-EONs[19]。由
式中:
4 SDM-EONs中串扰感知的虚拟网络映射算法
虚拟网络映射算法分为虚拟节点优先映射算法与虚拟节点和链路同时映射算法,但是虚拟节点和链路同时映射算法的时间复杂度非常高,所以本文采用虚拟节点优先映射算法,该算法先映射所有虚拟节点,然后映射虚拟链路。虚拟节点映射:当一个虚拟网络请求到达时,对虚拟节点和物理节点的优先级进行排序,实现一一对应映射。首先,利用虚拟节点邻近排序方法对虚拟节点的优先级进行排序,根据此排序方法寻找候选物理节点;然后,在选择候选物理节点的过程中,利用物理节点度分类方法把物理节点分为两类,并且根据所提出的物理节点优先级评估方法进行物理节点排序。虚拟链路映射:对虚拟链路和物理链路的优先级进行排序并且进行纤芯选择和频谱分配,最后完成虚拟网络映射。本文根据虚拟链路带宽大小进行优先级排序,基于此排序使用K最短路径算法寻找K条最短跳光路并使用所提出的物理链路优先级方法对候选物理链路排序。
4.1 虚拟节点邻近排序方法
在映射虚拟节点时,需要评估所有虚拟节点映射的优先程度,虚拟节点权重大则优先级高,且优先级高的虚拟节点先进行映射。而虚拟节点目前所能考量的属性为节点度、计算资源和节点相邻链路的带宽,故本实验参考文献[20]中虚拟节点权重
式中:
4.2 物理节点度分类方法
在第一步筛选候选物理节点过程中,若要提高虚拟网络映射率,需要满足以下4个条件:待映射物理节点的计算资源大于虚拟节点所需资源、虚拟节点所映射的物理节点度大于或等于虚拟节点度、未被同一虚拟光网络中的其他虚拟节点使用、此物理节点的相连物理链路上连续频谱隙数超过虚拟链路所需的频隙数,可以提高虚拟网络映射率。如
4.3 物理节点优先级评估方法
现有的物理节点优先级评估方法[10-11]考虑了物理节点的计算资源和物理节点相邻链路带宽资源碎片化程度,但是没有考虑与候选物理节点相连的物理链路上的芯间串扰,在虚拟链路频谱分配时,芯间串扰超过串扰阈值会使虚拟链路映射失败,最终导致虚拟网络映射失败。因此在选择候选物理节点时,应将其相连物理链路上串扰影响评估结果作为重要参考指标,均衡链路上的芯间串扰,以达到削弱全网的芯间串扰效果并且提高虚拟网络映射接受率的目的。定义候选物理节点优先级
式中:
4.4 物理链路优先级评估方法
当链路带宽资源的评估方式只考虑链路带宽资源大小时,会出现链路带宽资源比较丰富,但是频谱碎片化程度较高、满足虚拟链路带宽需求的可用频谱块数量较少的情况。因此,定义候选工作光路的路径优先级
式中:
4.5 纤芯均衡负载方法
为了使各个业务带宽在每根纤芯上均匀分布,均衡纤芯负载,提高虚拟网络映射率,提出一种纤芯均衡负载方法。首先,根据以下公式计算得到每根纤芯平均负载的带宽频隙数
式中:
4.6 频谱分区分配方法
确定虚拟链路映射的候选光路后,采用图顶点着色理论[21]将不相邻的纤芯分为一组,如
若虚拟业务在所应分配纤芯的S1~S3区域没有空闲频谱块可供分配,就在S4区域寻找;如果未寻找到满足虚拟业务的空闲频谱块,则依次遍历该纤芯所在的纤芯组中的S4区域和公共区域;如还未有满足需求的空闲频谱块,则遍历其他纤芯组的S4区域和公共区域,直至寻找到满足虚拟链路需求的频谱块并且所分配频隙产生的串扰影响小于串扰阈值,否则映射失败。这样的频谱区域分配方法可以减小芯间串扰,同时提高频谱利用率,最后达到降低带宽阻塞率的效果。
4.7 算法描述
CA-VNM算法的映射过程如下:
1)虚拟节点映射阶段:当一个虚拟网络请求
2)虚拟链路映射阶段:对于虚拟网络请求
CA-VNM算法的目的是在虚拟网络映射过程中动态评估芯间串扰、减小芯间串扰以及提高网络接收率。CA-VNM算法的执行过程如
CA-VNM的时间复杂度主要来源于3个部分:虚拟节点排序、物理节点排序和路由频谱分配。经过算法分析,虚拟节点排序的复杂度为
5 仿真验证与结果分析
5.1 实验环境配置及评价指标
为了验证SDM-EONs中串扰感知算法的性能,仿真使用Python语言在PyCharm2020.2.3平台上进行仿真。在
物理网络中每个物理节点包含400个计算资源。虚拟光网络请求到达服从参数为
仿真指标为虚拟网络请求接受率(VRR)、频谱利用率(SU)以及芯间串扰改善率(XTIR)。其中,虚拟网络请求接受率
式中:
频谱利用率
式中:
芯间串扰改善率
式中:
仿真的对比算法选取LCLC算法[11]、SD-RCSA算法[14]和文献[20]中的虚拟节点映射算法。这3种算法都没有采用SDM-EONs下的虚拟网络映射,因此将SDM-EONs下的虚拟网络映射记为LCLC-VNM、SD-VNM和SVNE-RMSC。同时,为验证所提虚拟节点映射方法的合理性,在LCLC算法的基础上增加了虚拟节点邻近排序方法、物理节点度分类方法和物理节点优先级评估准则,命名为LCLC-ND算法;为验证虚拟带宽请求分配的合理性,在LCLC算法的基础上增加纤芯均衡方法和频谱区域分配方法,命名为LCLC-FS算法。为了保证仿真的准确性,选取虚拟网络到达数为5000,并且每组仿真进行三次,取其平均值作为最终结果。
5.2 仿真结果与分析
表 2. 不同算法的时间复杂度对比
Table 2. Comparison of time complexity for different algorithms
|
图 7. NSFNET 网络中虚拟网络请求接受率
Fig. 7. Virtual network request acceptance rate in NSFNET network
6 结论
提出SDM-EONs中串扰感知的虚拟网络映射算法,探析基于SDM-EONs中虚拟网络映射的芯间串扰约束条件。在评估虚拟节点优先级时考虑虚拟节点之间的邻近性,提出虚拟节点邻近排序方法;在选择候选物理节点时,考虑物理节点度,使更多的底层物理资源提供需求更大的虚拟请求业务。此外,提出串扰影响评估方法,在此基础上评估物理节点的权重,平衡芯间串扰。在纤芯选择时,根据虚拟链路所需带宽大小进行纤芯选择,均衡纤芯负载。最后对频谱进行区域划分,并且在分配频谱时交替使用首次命中和最后命中方式,以减小芯间串扰影响。仿真结果表明,与LCLC-VNM算法相比,在网络负载较小时,所提算法的虚拟网络接收率最高优化了11%,频谱利用率最高提升了17.1%,芯间串扰改善率最好能达到67.7%,可见所提算法有利于提高虚拟网络的接受率,并且能有效改善SDM-EONs中的芯间串扰,达到优化虚拟网络映射性能的效果。
[1] 刘彤彤, 何荣希, 陈晓静. 流量疏导SD-EON中的持续时间感知多径路由与频谱分配算法[J]. 激光与光电子学进展, 2020, 57(21): 210602.
Liu T T, He R X, Chen X J. Holding-time-aware multipath routing and spectrum allocation algorithm in SD-EON with traffic grooming[J]. Laser&Optoelectronics Progress, 2020, 57(21): 210602.
[2] 刘岩, 何荣希, 王钜霖, 等. 弹性光网络中面向立即预留请求的碎片感知路由频谱分配算法[J]. 激光与光电子学进展, 2022, 59(5): 0506008.
Liu Y, He R X, Wang J L, et al. Fragmentation-aware routing and spectrum allocation algorithm for immediate reservation requests in elastic optical networks[J]. Laser&Optoelectronics Progress, 2022, 59(5): 0506008.
[3] Essiambre R J, Kramer G, Winzer P J, et al. Capacity limits of optical fiber networks[J]. Journal of Lightwave Technology, 2010, 28(4): 662-701.
[4] Muhammad A, Zervas G, Forchheimer R. Resource allocation for space-division multiplexing: optical white box versus optical black box networking[J]. Journal of Lightwave Technology, 2015, 33(23): 4928-4941.
[5] Zhang P Y, Yao H P, Liu Y J. Virtual network embedding based on computing, network, and storage resource constraints[J]. IEEE Internet of Things Journal, 2018, 5(5): 3298-3304.
[6] Zhu R J, Zhao Y L, Yang H, et al. Dynamic virtual optical network embedding in spectral and spatial domains over elastic optical networks with multicore fibers[J]. Optical Engineering, 2016, 55(8): 086108.
[7] 涂佳静, 李朝晖. 空分复用光纤研究综述[J]. 光学学报, 2021, 41(1): 0106003.
[8] Xuan H J, Wang Y P, Xu Z Q, et al. Virtual optical network mapping and core allocation in elastic optical networks using multi-core fibers[J]. Optics Communications, 2017, 402: 26-35.
[9] Chen Q, Chen B W, Zheng W W, et al. Virtual optical network mapping approaches in space-division-multiplexing elastic optical data center networks[J]. Journal of Lightwave Technology, 2022, 40(12): 3515-3529.
[10] 刘焕淋, 胡会霞, 马敬, 等. 光路可靠性和频谱整合因子感知的虚拟光网络生存性映射[J]. 电子学报, 2021, 49(4): 744-749.
Liu H L, Hu H X, Ma J, et al. Survivable embedding of virtual optical network based on lightpath reliability and spectrum integration factor-aware[J]. Acta Electronica Sinica, 2021, 49(4): 744-749.
[11] Chen B W, Zhang J, Xie W S, et al. Cost-effective survivable virtual optical network mapping in flexible bandwidth optical networks[J]. Journal of Lightwave Technology, 2016, 34(10): 2398-2412.
[12] Liu S X, Cheng B K, Hang C, et al. Combined routing and core-spectrum assignment scheme based on spectrum status for spatial division multiplexing elastic optical networks[J]. Proceedings of SPIE, 2020, 11435: 114350I.
[13] Zhang J, Yao Q Y, Bao B W, et al. Resource-oriented RMCSA scheme with low crosstalk effect in multi-core fiber-based elastic optical networks[J]. Optical Fiber Technology, 2022, 68: 102796.
[14] 刘仕鑫. 空分复用弹性光网络中路由纤芯和频谱分配策略及其应用的研究[D]. 南京: 南京邮电大学, 2020: 12-16.
LiuS X. Research on routing and core-spectrum assignment scheme and application in spatial division multiplexing elastic optical networks[D]. Nanjing: Nanjing University of Posts and Telecommunications, 2020: 12-16.
[15] Ren D P, Wang W, Hu J H, et al. Survivable virtual optical network coordinated mapping with local backup[J]. Journal of Optical Communications, 2021, 42(2): 265-271.
[16] 宣贺君, 王宇平, 徐展琦, 等. 多纤芯弹性光网络中纤芯选择算法[J]. 光学学报, 2016, 36(12): 1206005.
[18] KlinkowskiM, WalkowiakK. Impact of crosstalk estimation methods on the performance of spectrally and spatially flexible optical networks[C]//2018 20th International Conference on Transparent Optical Networks (ICTON), July 1-5, 2018, Bucharest, Romania. New York: IEEE Press, 2018.
[19] Zhu J, Zhu Z Q. Physical-layer security in MCF-based SDM-EONs: would crosstalk-aware service provisioning be good enough?[J]. Journal of Lightwave Technology, 2017, 35(22): 4826-4837.
[20] Liu H L, Ma J, Chen Y, et al. A survivable VON embedding algorithm based on resource mean and spectrum coherence-aware in elastic optical networks[J]. Optical Fiber Technology, 2020, 54: 102103.
[21] MoritaK, HirataK. Dynamic spectrum allocation method for reducing crosstalk in multi-core fiber networks[C]//2017 International Conference on Information Networking (ICOIN), January 11-13, 2017, Da Nang, Vietnam. New York: IEEE Press, 2017: 686-688.
[22] Zhao J J, Bao B W, Yang H, et al. Holding-time- and impairment-aware shared spectrum allocation in mixed-line-rate elastic optical networks[J]. Journal of Optical Communications and Networking, 2019, 11(6): 322-332.
Article Outline
任丹萍, 张黎, 胡劲华. 空分复用弹性光网络中串扰感知的虚拟网络映射算法[J]. 光学学报, 2023, 43(5): 0506003. Danping Ren, Li Zhang, Jinhua Hu. Crosstalk-Aware Virtual Network Mapping Algorithm in Space Division Multiplexing Elastic Optical Networks[J]. Acta Optica Sinica, 2023, 43(5): 0506003.