WDM网络中IP多粒度流量疏导新路由算法

169 浏览量更新于2024-08-27 收藏 1.95MB PDF 举报

本文是一篇发表在Elsevier出版社期刊上的研究论文，标题为"一种新的多粒度流量疏导路由算法在IP over WDM网络中的应用"。该研究主要针对的是光网络（WDM，Wavelength Division Multiplexing）与互联网协议（IP）融合环境下的网络设计问题。在现代通信系统中，随着数据流量的增长和带宽需求的多样化，如何有效地管理IP流量并确保其高效传输成为了关键挑战。作者们，来自东北大学信息科学与工程学院的Weigang Hou、香港理工大学计算机科学系的Lei Guo以及Xingwei Wang，提出了一种创新的多粒度流量疏导路由算法。这个算法旨在通过精细控制不同级别的流量颗粒，如包、段或流，来优化网络资源的分配和利用，从而提高网络的整体性能和效率。传统的路由策略往往集中在单个层次，而这种新的多粒度方法考虑了流量的复杂性，能够更灵活地应对流量的动态变化。在WDM网络中，这种算法可能会涉及到波长资源的调度、带宽预留和路径选择等多个层面的操作。它可能采用了先进的流量预测技术，以便提前规划并适应突发的流量需求。此外，文章可能还探讨了算法的实现细节，包括如何处理网络拓扑变化、如何处理拥塞和故障恢复等问题。为了确保研究成果的合法使用，文章指出这只是一个供作者内部非商业研究和教育用途的版本，不允许用于复制、分发、销售或许可给第三方网站。作者们可以将他们的论文版本上传到个人网站或机构仓库，但必须遵守Elsevier的版权政策和存档手册，详情可访问http://www.elsevier.com/copyright。这篇论文为IP over WDM网络的流量管理提供了一个新颖且具有实践价值的解决方案，对于网络运营商和研究人员来说，它可能成为理解和优化未来网络架构的关键参考文献。

Author's personal copy

1020 W. Hou et al. / Optik 122 (2011) 1019–1029

1.1. Related work

With the number of wavelengths in a ﬁber increases, ordinary

OXCs that only switch wavelength-granularity demands require

a large number of transmitting ports, which results in increasing

cost and control complexity. MG-OXC is the foremost solution due

to its small scale modularity, crosstalk and complexity reduction.

The authors in Refs. [6,7] proposed the use of MG-OXC without

any other additional function matrixes. But recently, the concept

of using sub-wavelength, wavelength, waveband and even ﬁber

switches in a hierarchical manner has received growing attention,

and then the authors in Refs. [8,9] provided MG-PXC that contains

a three-layer MG-OXC with the function of converting wavelength

or waveband and a OEO trafﬁc grooming unit used to support mul-

tiplexing LRSs. Since the joint routing algorithm and the integrated

grooming policy based on MG-PXC was not considered in Ref. [9],

the proposed MG-PXC was lack of the necessary policy control

plane and improvements for decreasing the overall cost of the IP

over WDM network system.

On the other hand, an integrated auxiliary graph was pre-

sented in Ref. [10] to compute the route and assign the wavelength

for lightpath synchronously with constrained transceivers in a

dynamic environment. Based on that integrated auxiliary graph,

the authors proposed an algorithm called IGA to obtain lower

blocking probability followed by multi-hop grooming since it can

perform full-wavelength conversion on each node. But the prob-

lem here is that, the full-wavelength conversion may consume a

large number of OEO ports inevitably. In addition, IGA and other

traditional trafﬁc grooming algorithms choose route for lightpath

in wavelength-routed manner. If the lightpath has no available

bandwidth for LRSs, we will compute another wavelength-route

for establishing new lightpath with two same end nodes, and that

new wavelength-route also consumes transmitting ports at inter-

mediate nodes, which is not desirable. Waveband switching can

save additional consumption of transmitting ports because it can

bind several wavelength-routes into a waveband tunnel as a sin-

gle entity. The WBS algorithms and several WM-strategies were

introduced in Refs. [11–17], among which, the authors in Ref. [16]

described a new dynamic WBS algorithm called RA-IAG followed

by sub-path merging strategy, compared to the other WBS algo-

rithms followed by end-to-end merging strategy, it can save more

OOO transmitting ports. But RA-IAG did not have the function of

trafﬁc grooming and also performed full-wavelength conversion

in each node. The additional consumption of OEO ports has not

still been solved. A practical way to solve this problem is to pay

attention on sparse placement of a limited number of wavelength

converters or limited-range wavelength converters. The authors in

Ref. [17] proposed a waveband switching algorithm with consid-

ering the intra-band wavelength conversion technology, where a

wavelength can only be converted to any other wavelengths within

the same waveband. Intra-band wavelength conversion technology

has an ability of reducing the times of converting wavelengths so

that it can save the consumed OEO ports. However, the authors in

Ref. [17] only considered the wavelength-granularity demands and

ignored the sub-wavelength-granularity demands, i.e., LRSs. More-

over, although the idea of integrating the advantages of traditional

trafﬁc grooming and waveband switching was ﬁrst presented in

Ref. [9], how to perform the multi-granularity trafﬁc grooming was

not described in this paper. Therefore, the multi-granularity trafﬁc

grooming is an interesting and challenging problem we face in IP

over WDM networks.

1.2. Our contribution

As mentioned above, some wavelength-convertible trafﬁc

grooming algorithms can attain low blocking probability by incor-

porating wavelength conversion capability at all of the nodes.

However, as the number of active wavelengths in a ﬁber increases,

the solution of multiplexing several LRSs into the high-capacity

lightpaths, whose consumed wavelengths are different from each

other, followed by full-wavelength conversion technology may

cost a large number of expensive OEO ports and also may require

many transceivers if most of lightpaths are set up anew since each

new lightpath will consume transceivers on its two end nodes.

The idea of implementing with the limited wavelength conversion

technology to save OEO ports by decreasing the times of wave-

length conversion is readily necessary. Therefore, the intra-band

wavelength conversion is conceivably of interest to us. In addi-

tion, the waveband switching wherein several wavelengths can be

grouped into a waveband and be switched as a single entity can save

more all optical transmitting ports than other technologies using

wavelength routing switches at intermediate nodes. Therefore, in

our opinion, the intra-band wavelength conversion and waveband

switching are both practical and should be considered when solving

the multi-granularity trafﬁc grooming problem.

In this paper, we devise an identiﬁed structure of MG-PXC

and the new integrated grooming policy (IGP) that contains three

WG-schemes and two WM-strategies, introduce three important

technologies including multi-hop grooming, intra-band wave-

length conversion and waveband switching, and present the

mathematical model for calculating the number of consumed ports.

Since the trafﬁc grooming problem is the NP-hard [18], we pro-

pose a new heuristic joint routing algorithm MGIAG for routing

trafﬁc with different granularities starting from sub-wavelength

granularity (i.e., LRSs) to individual wavelength granularity and

then to waveband granularity. Finally, to support our algorithm

MGIAG, we employ a novel graph model called IAG. The algorithm

MGIAG is developed based on the auxiliary graph IAG and applies

the least-cost-path routing algorithm in that auxiliary graph for

each trafﬁc demand and updates that auxiliary graph accordingly.

Our proposed MGIAG can integrate the advantages of traditional

trafﬁc grooming using multi-hop grooming scheme and waveband

switching using sub-path waveband merging strategy effectively to

save OOO ports and obtain low blocking probability. On the other

hand, MGIAG can reduce a large number of OEO ports by using the

intra-band wavelength conversion.

This paper is organized as follows. In Section 2, we ﬁrstly intro-

duce the identiﬁed structure of MG-PXC, the multi-hop grooming,

the intra-band wavelength conversion and the waveband switch-

ing technologies in our proposed mechanism, and then we present

the integrated grooming policy and demonstrate how to calculate

consumed ports by using our proposed mathematical model. In

Section 3, we construct an integrated auxiliary graph according to

the network state. Based on that auxiliary graph, we proposed the

joint routing algorithm MGIAG in Section 4, and give an illustrative

example to explain the routing and wavelength/waveband assign-

ment in MGIAG in Section 5. In Section 6, we make the simulation

and performance analysis. Section 7 concludes this paper.

2. Problem statement

2.1. Network model

The physical network topology can be denoted as PNT (N, L,

B, W, T, G), where N is the set of physical nodes, L is the set

of ﬁber links, B is the set of active wavebands per ﬁber link, W

is the set of active wavelengths per ﬁber link (i.e., ﬁber capac-

ity), T is the set of tunable transceivers per physical node and

G is waveband granularity, in this work, we ensure



G. The set of wavelengths in a waveband is speciﬁed before-

hand. Each waveband B

(1 ≤ y ≤



) consists of G contiguous

剩余11页未读，继续阅读

weixin_38681628

粉丝: 4

WDM网络中IP多粒度流量疏导新路由算法

WDM网络中基于流量划分的多粒度修饰新算法

IP over WDM网络中解决流量分集的多粒度修饰新算法

多纤IPover WDM网中的一种新型联合路由算法 (2006年)

WDM网络中基于优先级的多任务波长路由分配算法

WDM光网络中的组播路由算法研究

WDM网络中一种高效的动态多播流量整理算法

IP /WDM网络中自适应区分权重的H TA动态业务流疏导算法 (2010年)

全光WDM网络中的动态波长路由算法 (2006年)

WDM网状网动态流量疏导算法研究

IP/WDM网络中的DWG-HTA动态业务流疏导算法

最新资源