Multiple―Constraint―Aware RWA Algorithms Based on a Comprehensive Evaluation Mod

开篇：润墨网以专业的文秘视角，为您筛选了一篇Multiple―Constraint―Aware RWA Algorithms Based on a Comprehensive Evaluation Mod范文，如需获取更多写作素材，在线客服老师一对一协助。欢迎您的阅读与分享！

Abstract

Because of explosive growth in Internet traffic and high complexity of heterogeneous networks， improving the routing and wavelength assignment （RWA） algorithm in underlying optical networks has become very important. Where there are multiple links between different the node pairs， a traditional wavelength-assignment algorithm may be invalid for a wavelength-switched optical networks （WSON） that has directional blocking constraints. Also， impairments in network nodes and subsequent degradation of optical signals may cause modulation failure in the optical network. In this paper， we propose an RWA algorithm based on a novel evaluation model for a WSON that has multiple constraints. The algorithm includes comprehensive evaluation model （CEM） and directional blocking constraint RWA based on CEM （DB-RWA）. Diverse constraints are abstracted into various constraint conditions in order to better assign routing and wavelength. We propose using the novel CEM to optimize routing according to an assessed value of constraints on transmission performance. This eliminates the effects of physical transmission impairments in a WSON. DB-RWA based on CEM abstracts directional blocking conditions in multiple links between network nodes into directional blocking constraints. It also satisfies rigorous network specifications and provides flexibility， scalability， and first-fit rate for the backbone， especially in multiple links between WSON nodes.

Keywords

RWA； WSON； multiple links between nodes pair； directional blocking constraint； comprehensive evaluation model

1 Introduction

The Internet has become a critical piece of infrastructure in modern societies， and explosive growth in the number of Internet users has given rise to a corresponding explosion in Internet traffic. Over time， data packets have become the main type of traffic being transmitted over networks. An optical network carries not only fixed-bandwidth voice services but also variable-bandwidth services， mobile services， and multimedia services. Optical networks must support grid computing， file downloading， video-on-demand， storage-area networks， and other new applications. Meanwhile， demand is growing for capacity， responsiveness， resilience， quality， and different types of services. Underlying optical networks are ultrahigh-capacity， large-scale platforms that support synchronous optical network and synchronous digital hierarchy （SONET/SDH）， multiprotocol label-switching transport profile （MPLS-TP）， and optical transport network （OTN） for transporting data at different granularities. In large-scale commercial applications， wavelength-switched optical networks （WSONs） must satisfy the needs of network service providers， who face strong pressure to increase availability， improve reliability， and reduce bandwidth cost.

Nodes in a WSON―unlike nodes in other types of generalized multiprotocol label-switching （GMPLS） networks―are highly asymmetric in their switching capabilities， and compatibility of signal types and network components needs to be considered. Label assignment can also be nonlocal [1]. In a WSON， wavelength continuity and resource availability constraints must be satisfied in order to determine optimal and cost-effective optical connections （paths） [1]. Such determining of paths is referred to as routing and wavelength assignment （RWA）.

To design a new WSON for complex contexts， multiple constraint conditions need to be applied to the evaluation model and RWA algorithms. For redundancy and reliability in optical networks， most network operators place multiple optical fibers between the network nodes， especially in the backbone network or main part of an access or metro network. Therefore， application scenarios based on multiple links between network nodes have become the focus of attention of network operators. Traditionally， RWA algorithms have been used for only one link between network nodes； however， such algorithms are invalid for routing between different links and nodes [2].

Because there are impairments of varying degrees in a large-scale optical network， degradation can cause unacceptable bit error rates or even a complete failure to detect and demodulate the received signal. Therefore， optical impairments need to be taken into account in an RWA algorithm so that signals with acceptable quality can be propagated in the WSON.

Many studies have been done on quality of transmission （QoT） compensation as well as impairment-aware routing and wavelength assignment （IA-rwa） [3]-[5] based on path computation element （PCE） strategies and extended resource reservation protocol （RSVP） [6]. Many traditional RWA algorithms have been studied only in one link field [7]， [8]. When there are multiple links between the network nodes， few traditional RWA algorithms can successfully compute the optimal path， and few have been evaluated in terms of impairment assessment model [9].

In this paper， we describe a novel RWA algorithm based on a model for assessing multiple constraints in a WSON. The algorithm comprises a comprehensive evaluation model （CEM） and directional blocking RWA based on the CEM （DB-RWA）. Diverse scenarios are abstracted into various constraint conditions so that routing and wavelength can be better assigned. We propose a novel CEM that optimizes routing based on a value derived from an assessment of constraints on transmission performance. The algorithm eliminates the effects of impairments in WSON transmission.

Directional blocking conditions are abstracted by DB-RWA based on CEM in multiple links between the network nodes into directional blocking constraints that satisfy the rigorous network specifications. DB-RWA based on CEM provides scalability and flexibility for the backbone， especially in multiple links between WSON nodes.

2 Comprehensive Evaluation Model

When an optical signal is being transmitted in a fiber link and is passing through an optical device or cell， physical impairments―such as reduced signal strength； triggering noise； or changing time domain， frequency domain， or polarization properties―are ineluctable. Especially in a transparent or translucent light network， which lacks photoelectric regenerators， the effects of impairments accumulate as the optical signal is transmitted. When an impairment threshold is exceeded， quality of transmission is seriously affected. Transmission impairments can be linear or nonlinear. The effects of linear impairments of every wavelength are independent， and the effects of nonlinear impairments are not independent. An IV module is used to estimate the transmission quality of the chosen wavelength path so that the path meets QoT requirements [10]-[12].

Currently， the evaluation model most widely used is a multitarget constraint evaluation model. In this model， constraints are considered individually， and a light path is only eligible when it satisfies the thresholds for each target constraint. However， because this model ignores the interaction of factors， the estimation may be highly inaccurate.

In a comprehensive evaluation model， physical impairment is reflected in the error rate， which reflects a variety of impairments. This model uses a simulation and analytical fitting method to comprehensively evaluate physical impairments. It also uses bit error rate （BER） as a constraint when evaluating the QoT of a light path. Tolerance security reserve can be obtained from the BER.

A comprehensive evaluation model takes into account the interaction of physical impairments. It requires much calculation； the model is complex； and much information is acquired. However， the accuracy of the evaluation is greatly improved. A general way of implementing the model is through analytical and experimental simulation （fitting）. By transforming the physical impairment into a BER， we can measure the quality of light transmission.

2.1 Sketch of Evaluation Model

BER is a key parameter for assessing systems that transmit digital data. BER is used to assess the full end-to-end performance of a system that includes transmitter， receiver， and the medium in-between. BER allows the performance of a whole system in operation to be tested， which differs from testing individual component parts and hoping that they will operate satisfactorily when in place. The BER is the rate at which errors occur in a transmission system. This can be directly translated into the number of errors that occur in a string of bits. According to the principle of Gaussian noise distribution， Q -factor corresponds to BER. By evaluating Q -factor， the quality of a transmission system can be estimated. Q -factor is given by

Where g （） is the accumulated cost of each kind of impairment in the transmission， and ζ （） is the model effect caused by individual impairments such as dispersion， cross-phase modulation （XPM）， and four-wave mixing （FWM）. In normal circumstances， ζ （）= 1 when amplified spontaneous emission （ASE） noise plays a main role. As nonlinear effects increase，

ζ（） decreases below 1. This formula for evaluating transmission quality takes into account each kind of physical impairment， including residual dispersion， nonlinear phase shift， ASE noise， and polarization mode dispersion （PMD） [13]-[15]. The Q-factor cost of each physical impairment can be obtained by a semianalytical method in order to obtain the total performance index.

OSNR ref is the reference of the optical signal-to-noise ratio （obtained by experiment） when the BER requirement is met， that is， when Q =Q ref. The reference OSNR ref is a function of physical impairments， including residual dispersion， nonlinear phase shift， PMD， and crosstalk. The Q function can be ・N is the number of the fiber links between adjacent nodes

・M is the number of wavelengths in every fiber

・i is the i th wavelength， i = 1， 2，...， M

・j is the j th fiber， j = 1， 2，...， N

・aj，i is the occupancy of the i th wavelength in the j th fiber， 0 = occupied， 1 = available

・bi is the number of available fibers with the i th wavelength

・ci is the occupancy of the i th wavelength. If the i th wavelength of any fiber is available， ci = 1. If the i th wavelength of every fiber is occupied， ci = 0.

We implemented and evaluated the proposed approach in the eighteen-node topology （Fig. 2）. The testbed has the same topology as the network topology， in which there are10 fiber links per hop， and each fiber link contains five wavelengths. In terms of traffic characteristics， uniformly distributed light path requests arrive at the network following a Poisson process. The holding time of the light path is exponentially distributed using unit mean.

For the directionless condition， if the traffic moves from node A to K， the path derived using D algorithm is A-F-G-K. If the paths are calculated using KSP algorithm， the first best path is A-F-G-K； the second best path is A-B-F-G-K； and the third best path is A-F-G-L-K. For the directional condition， we assume that link A-F and link B-F can only exchange with link F-E. The three best paths cannot meet the directional limitation， but the path A-F-E-H-G-K can be used. As a result， we can try to find the RWA solution in the directional condition.

4 Simulation and Results

4.1 CEM Experimental Simulation System

Because an analytical solution to OSNR cost of impairment cannot be obtained directly， a simulation or experimental fitting can be used to work out a second cost curve of impairment OSNR and obtain an approximate analytical solution.

In this simulation method， the structure of the simulation system in Fig. 3 can be used.

At the transmitting terminal， the required number of wavelengths can be dispersed to emulate the physical impairment in the context of a single channel and multiple channels. After the optical multiplexer， the optical signal enters a dispersion compensating fiber （DCF） with predispersion compensation. Predispersion compensation depends on the value of dispersion parameters， span residual dispersion， and number of spans. Postdispersion compensation depends on the residual dispersion configuration of a span. The receiver performs decoding and judging based on the demux and demodulator and arrives at the BER.

In the simulation of OSNR cost curves， only one kind of impairment should generally be considered， and other impairments should be ignored. Other system parameters should be fixed. In simulation， parameters for the transmission fiber span， transmitting signal， in-span DCF， pre-DCF， and post-DCF are fixed. The fixed parameters of the transmission fiber span include fiber type， optical attenuation coefficient， optical fiber dispersion coefficient， span length， and nonlinear coupling coefficient. The fixed parameters of the transmitting signal are modulation format， signal rate， signal transmission power， symbol line code， duty cycle of optical pulse， rolling pressure coefficient， channel interval， and center wavelength. In-span DCF contains dispersion coefficient， attenuation， and nonlinear coupling coefficient. The length of the compensation fiber depends on each segment’s residual dispersion value. Pre-DCF and post-DCF maintain the dispersion coefficient， nonlinear coupling coefficient， and fiber length （which is determined by compensation value）.

When calculating the impairment OSNR cost curve， we can change the number of spans and the scheme for the span dispersion compensation in order to obtain more comprehensive data.

4.2 DB-RWA-Based CEM Simulation

and Results

We have implemented and evaluated the proposed approach in the eighteen-node topology （Fig. 2）. The testbed describes the same topology as the network topology in

Fig. 2， in which 10 fiber links per hop are assumed. Each fiber link contains five wavelengths. In terms of traffic characteristics， uniformly distributed light path requests arrive at the network following a Poisson process. The holding time of light path is exponentially distributed using unit mean.

In the directionless condition， the simulation environment is independent of the direction limitation. The statistics are obtained for 100 to 1500 Erlangs. Table 2 shows the blocking rates and setup delays.

Fig. 4（a） is based on the data in Table 2.

We assume the leaving rate μ = 0.1 and the arriving rate ranges from 10 to 80services per second. The statistics are obtained for 100 to 800 Erlangs. The blocking rate and the setup delay are shown in Table 3.

Fig. 4（b） is based on the data in Table 3. As the traffic load increases， blocking probability increases. Fig. 4（b） also confirms that the simulation is correct and reasonable. The setup delay is around 0.457 ms with small fluctuation. The setup delay is calculated from the beginning of route calculation to the end of wavelength assignment. Because the K of the KSP algorithm remains unchanged， the setup delay barely changes in different arriving-rate scenarios.

In the directional condition， where the network is limited by direction， we assume the arriving rate remains at 40 services per second and the leaving rate remains 0.1. KSP algorithm followed by the K-value is equal to 2， 3， ...， 8. Table 4 and Fig. 5 show the blocking probabilities and the setup delays.

Fig. 5 shows the setup delays with K values. When K increases， the setup delay， which is determined by the complexity of the KSP algorithm， gradually becomes longer. According to the data in Table 4， K increases， the probability of blocking reduction decreases. When the number of alternative paths increases， there is more chance a route will be chosen that has a lower chance of blocking.

5 Conclusion

In this paper， we have presented a novel CEM and DB-RWA algorithm based on CEM for use in a WSON with multiple constraints. Taking into account the interaction between physical impairments， a comprehensive evaluation model eliminates the effects of physical transmission impairments in a WSON. CEM improves the evaluation accuracy in the mathematical method and reasonably obtains and calculates large amount of information. The implementation method and quality of light transmission are analytical and transform the physical impairment factor into a BER price， especially in the transparent light network that lacks photoelectric regenerators. DB-RWA algorithm based on CEM in multiple constraint conditions in WSON not only solves the problem of RWA in multiple links between network adjacent nodes but also substantially decreases the setup delay and cost of computation. With increasing scope in the multiple links between adjacent nodes in the network， time performance can be optimized. The simulations in both directional and directionless condition demonstrate that the network is optimized to for shorter routing time as the number of network nodes and links in adjacent nodes increases. An optimal path can be chosen for ultrahigh capacity with lower blocking probability.

Acknowledgment

This work was supported in part by 973 Program （2010CB328204）， NSFC project （60932004） and RFDP Project （20090005110013）.

References

[1] Framework for GMPLS and Path Computation Element （PCE） Control of Wavelength Switched Optical Networks （WSONs）， IEEE RFC 6163， Apr. 2011.

[2] Xiaowen Chu， Bo Li， I. Chlamtac， “Wavelength converter placement under different RWA algorithms in wavelength-routed all-optical networks，” Communications， IEEE Trans. Commun.， vol. 51， no. 4， pp. 607-617， Apr. 2003.

[3] S. Azodolmolky， Y. Pointurier， M. Angelou， D. Careglio， J. Sole-Pareta， I. Tomkos， “A Novel Impairment Aware RWA Algorithm With Consideration of QoT Estimation Inaccuracy，” in IEEE/OSA J. Opt. Commun. Netw.， vol. 3， no. 4， pp.290-299， Apr. 2011.

[4] D. Adami， S. Giordano， M. Pagano， L. Zuliani， “Online light path provisioning and critical services： New IA-RWA algorithms to assure QoT and survivability，” in Proc. 12th Int. High Performance Switching and Routing Conf.， Cartegna， Spain， Jul. 2011.， pp. 101-106.

[5] S. Azodolmolky， Y. Pointurier， M. Klinkowski， E. Marin， D. Careglio， J. Sole-Pareta， M. Angelou， I. Tomkos， “On the Offline Physical Layer Impairment Aware RWA algorithms in transparent optical networks： State-of-the-art and beyond，” in Proc. 13th Opt. Netw. Design and Modeling （ONDM’09）， Braunschweig， Germany Feb. 2009.

[6] Ying Wang， Jie Zhang， Yongli Zhao， XuPing Cao， Hai Huang， WanyiGu， YuefengJi， “Double Cooperating PCEs for IA-RWA computation with QoT compensation in transparent optical networks，” in Proc.36th European Conf. Opt. Commun.， （ECOC’10）， Sep. 2010， pp. 1-3.

[7] K. Manousakis， K. Christodoulopoulos， E. Varvarigos， “Impairment-Aware Offline RWA for Transparent Optical Networks，” in Proc. INFOCOM 2009， Rio de Janiero， Apr. 2009.

[8] D. Adami， S. Giordano， M. Pagano， L. Zuliani， “Online light path provisioning and critical services： New IA-RWA algorithms to assure QoT and survivability，” in Proc. 12th Int. Conf. High Performance Switching and Routing （HPSR’11）， Cartagena， Spain， pp.101-106.

[9] Yongtao Gong， Peiyuan Lee， WanyiGu， “A novel adaptive RWA algorithm in wavelength-routed network，” in Proc. Global Telecommunications Conf.（GLOBECOM'03）， San Fransisco， CA， vol. 5， pp. 2580-2584.

[10] J. Antona， S. Bigo， "Physical design and performance estimation of heterogeneous optical transmission systems，" in Comptes Rendus Physique， vol. 9， no. 9， pp. 963-984， 2008.

[11] A. Morea， N. Brogard et al.， “QoT function and A* routing： an optimized combination for connection search in translucent networks，” in J. Opt. Netw.， vol. 7， no. 1， pp. 42-61， 2008.

[12] F. Leplingard， A. Morea， T. Zami， N. Brogard， “Interest of an adaptive margin for the quality of transmission estimation for light path establishment，” in Proc. Conf. on Opt. Fiber Commun. （OFC’09）， San Diego， CA， paper OWI6， 2009.

[13] H. Keang-Po， “Error probability of DPSK signals with cross-phase modulation induced nonlinear phase noise，” in IEEE J. Selected Topics in Quantum Electronics， vol. 10， no. 2， pp. 421-427， 2004.

[14] Alberto Bononi et.al， “Cross-phase modulation induced by OOK channels on higher-rate DQPSK and coherent QPSK channels，” in J. Lightw. Tech.， vol.27， no. 18， pp. 3974-3983， 2009.

[15] Wenlu Chen， Shan Zhong， Zhonghua Zhu， Wei Chen， Chen， Y.-J.R.， “Adding OSNR monitoring functionality on AWG based power monitoring circuits，” in Opt. Fiber Commun. Conf. （OFC’02）， Anaheim， CA， pp. 668-670.

Manuscript received： January 11， 2012

Biographies

Hui Yang （） is a PhD candidate in communication and information systems at Beijing University of Posts and Telecommunications （BUPT）. His research interests include network architecture， dynamic optical network， flexible bandwidth optical network， software driven networks， and cross-stratum optimization. In 2011， he authored an academic paper about network architecture that was accepted by the Technical Program Committee of the Optical Fiber Communications/ National Fiber Optic Engineering Conference （OFC/NFOEC）.

Yongli Zhao （） is a lecturer at the Institute of Information Photonics and Optical Communications at BUPT. He received his BSc degree in communication engineering and PhD degree in electromagnetic field and microwave technology from BUPT in 2005 and 2010. He has published more than 80 articles. His research interests include wavelength switched optical networks， spectrum-efficient optical transport networks， and packet transport networks.

Shanguo Huang （） is an associate professor and doctoral supervisor at BUPT. He has presided over and participated in more than 10 national and provincial research projects， including the National Natural Science Fund， the Project of "863"， and the Key Plan and Military Projects of the Ministry of Education. He has also participated in projects of important research institutions and enterprises. He has authored more than 50 articles for international conferences and periodicals， 30 of which are indexed in SCI and EI. He has also authored three international guest reports， 18 patents for invention， three IETF international standard drafts， and two monographs. His main research interests include data optical network and ASON technology， and transmission network planning and optimization.

Dajiang Wang received his master’s degree in Machinery Manufacturing and Automation from Shanghai University in 2001. He was a system engineer and pre-research project manager for ZXUCP A200， and now is the planning manager for ZXUCP A200 products. He led several R&D projects on RWA algorithm， PCE， and unified control plane.

Xuping Cao （.cn） received his PhD from the Institute of Information Photonics and Optical Communications， BUPT. His research interests include next-generation optical networks. From 2004 to 2006， he worked in the optical network product line at Huawei. Since 2010， he has worked in the Advanced Optical Network Research Department at ZTE Corporation， where he researches MPLS/GMPLS network control plane technology. He has published more than 10 articles （indexed in SCI and EI）， 18 patents， and three IETF international standard drafts.

Xuefeng Lin （.cn） received his PhD degree in computer science from NorthEast University， China. He is currently a system engineer of Bearer Network Pre-Research Department， ZTE Corporation. His research interests include contral plane of optical transport network， and the integration between data equipment and optical equipment.