Development of an Algorithm for Physical Layer Impairment Aware Routing in WDM Network

Abstract

A communication system is an entity involving the set of physical equipments meant for information transfer from one place to another. Till 1980s, most of the communication systems implemented electrical transmission technology via copper cables, radio frequency, microwave and satellite links. The invent of lasers stirred focus of research towards implementing optical region of electromagnetic spectrum instead of High Frequency (HF) or Ultra High Frequencies (UHF) as carrier wave. The advantages like long transmission distance, large information carrying capacity, small size, immunity to electrical interference and signal security played a vital role in replacement of copper cable as communication medium. An important breakthrough came with invent of Wavelength Division Multiplexed (WDM) networks further supported by introduction of Erbium Doped Fiber Amplifiers (EDFA). The popularity of computer usage and internet applications resulted in exponential increase in data traffic transferred using optical networks. Recent trend in growth of internet traffic shows that voice traffic being transferred increase approximately 7% per year, while data traffic has been growing at rate much higher. To achieve these higher capacities, many new technological developments were done in the construction of lasers and fibers, installation procedures, network test and monitoring equipments. Implementation of WDM networks having dynamic network provisioning capabilities is sought as one of the possible solutions. WDM network involves modulation and multiplexing of optical carriers of different wavelengths into a fiber for transmission over long distances and has enormously increased the data carrying ability of the network. Theoretically, if the full transmission capacity of WDM be explored, bandwidth of tens of terahertz may be easily achieved. As the infrastructure of present system may easily handle 160 channels over the same fiber, a simple 10 Gbps signal may easily be expanded up to 1.6 Tbps. In the subsequent years, a lot of developments have taken place in optical networks to develop and deploy sophisticated and reliable communication system with focus on improving transmission fidelity, increased data rate or carrying capacity of the network, increased transmission distance between inline repeater or amplifiers. The deployment of optical components like Optical Add Drop Multiplexers (OADM), Optical Cross Connects (OXCs,) wavelength convertor switches etc on network nodes has begun the phase of transition in optical networks where requirement of Optical to Electrical (O-E) signal transfer during the data transmission from a source to destination point is minimized. Based on O-E conversions taking place in the optical links, WDM networks may be broadly categorized as opaque, translucent and transparent networks. A single optical hop of lightpath never cross more than one physical fiber link in opaque network, could cross one or more fiber links in translucent networks and cross the entire source-destination route in translucent networks. Although opaque networks makes the process of design, control and management easier, but are not the practical solution to increased traffic growth as there is a huge potential for reducing power consumption and cost by eradicating unnecessary opto-electronic (OE) transformations on a signal path in a core optical network. Transparent networks have advantages like data format and bit rate transparency, significant footprint and power savings, reduced network costs. A complete transition of backbone networks from opaque to transparent is a far off dream till date due to limitations like absence of wavelength conversions, lack of signal quality enhancement at each node, routing and restoration issues, interoperability among the multiple service providers etc. The networks where an O-E-O operation is performed at certain nodes, in accordance with the optical components used, are called translucent networks. The current generation point to point network architectures are evolving from traditional opaque networks towards translucent and transparent networks and is attracting the focus of research community for technological advancements. The dominating issues that need to be addressed while planning and operation of transparent and translucent networks include connection provisioning with focus on minimizing the network cost and improve the resource utilization efficiency and addressed in the present research work. The process of connection provision involves setting up lightpath for individual call requests and assigning wavelength to each connection and is known as routing and wavelength assignment (RWA). For each connection request, once the route is calculated considering various constraints of network topology, the lightpath is set-up by allocating specific wavelength to the signal. The quality of signal received at the destination node must satisfy the conditions laid by Service Level Agreements (SLAs) for different class of users. Most of the routing algorithms consider the optical layer to be the ideal one, thus making all the outcomes of routing algorithms valid and feasible. Practically, some of the calculated lightpath may not be satisfying the SLAs as the quality of signal transmission is influenced by the physical properties and operating conditions of optical fiber. Lack of O-E conversions taking place, plenty of network components involved, additive nature of Physical Layer Impairments (PLIs) and dynamic configuration of network further contribute towards the signal degradation. Accordingly, it is important for network designers and operators to know various important factors influencing the signal quality, techniques to monitor, model and mitigate their effects, various techniques to communicate the information regarding these disparaging effects to network layer and control plane protocols; and finally, how to use all these techniques in conjunction with control and management plane protocols to dynamically set up and manage optically feasible lightpath. In the present research work an effort has been put in to identify the different parameters affecting the signal quality which necessarily need to be monitored and controlled to meet the requirements of Quality of Service (QoS) and Service Level Agreement (SLA). A novel algorithm is developed for routing of optical networks considering the degradation effect posed by physical layer impairments. The regenerator placement and utilization is considered for optimum utilization of network resources and minimizing the blocking probability. The focus of algorithms is on comprehensiveness, transparency, fault localization capability and scalability of the system. The analysis and evaluation of the physical layer parameters affecting the signal quality is done using network simulations. The result of performance monitoring is used for the routing. The signal measurements done are used for network management with routing capabilities, where priority traffic that requires high performance can be dynamically tuned to the appropriate channels. The proposed system is cost effective with sufficiently high update speed and reduced processing time. In this work, different optical impairments responsible for signal degradation and their mathematical models, including higher order impairments, are studied. The different physical layer impairments are broadly classified as linear (Dispersion, attenuation) and non-linear impairments (Self Phase Modulation (SPM), Cross Phase Modulation (XPM), Four Wave Mixing (FWM)). The influence of these impairments on system performance in form of BER and Q-factor, influence of variations in system parameters like wavelength, length of fiber and interaction between different impairments is studied. The different techniques that may be implemented for monitoring are broadly categorized as digital and analog techniques [35]. The digital monitoring involves BER, Q-factor monitoring by studying the sequence of bits being transferred. The analog techniques involves polarization domain analysis using polarization nulling, or degree of polarization including [37-42], time domain analysis using eye diagrams, delay tap plots, constellation diagrams, histograms etc [43-61] and frequency domain analysis using optical or RF spectrum [62-68]. Digital techniques focus on digital bits being transferred through the fiber while analog techniques analyze the signals at the output to extract the information about network stage, the identification, location and measurement of different impairments being present. In the present work, the individual impairments are diagnosed through eye diagrams and optical spectrum, while the overall status of the network in presence of all the impairments is presented through BER and Q-factor. BER is used in Q-matrix of routing algorithm. Q-value is also measured to ensure the signal satifying the threshold conditions as per SLAs. The Q-value for signal propagating in the fiber link decreases with increased length of fiber due to higher accumulation of impairments in the fiber. The corresponding value of BER is increased. This is the reason making uniform placement of regenerators a neccesasity in optical fiber links. The quality of link is also influenced by the number of wavelengths propagating inside the optical fiber. The numerous heuristic algorithms are presented in the literature on finding the optimal solutions to RWA considering the role of physical layer on transmission. The impact may be inculcated either by considering impairments as one of the constraints to route calculations, known as Impairment Constraint Based Routing (ICBR), or by setting some threshold for quality factor of link and calculating route considering the links satisfying the minimum threshold, known as Physical Layer Impairment Aware Routing (PLIAR). These routing algorithms present either single impairment, multiple impairments (i.e. combination of few e.g. PMD, XPM or PMD, ASE, XPM, FWM) quantized in the form of either of BER, Q-factor, OSNR or aggregate power flowing in the links [71-88]. Most of the studies are based on assumption that the PLIs are of static nature therefore may be estimated analytically. But impairments like crosstalk, SPM, XPM, FWM depends upon the network state. Their values may vary by bare minimum variations in properties of signal being transmitted in channel e.g. power, bit rate or by the state of signal flowing in adjacent channels. In dynamically reconfigurable networks, there is a need to monitor them online and broadcast the information into the control plane through centralized or distributed architectures. In the reports demonstrating online monitoring of PLIs, only one or two impairments are considered. For accurate estimation of optical reach in links, it is necessary that not only all the impairments (both linear and non-linear) must be monitored, but the impact of mutual interaction of specific impairment must also be considered during RWA. In the present work, a routing algorithm is presented which take into account the impact of impairments on the quality of signal flowing in the links. The routing algorithm assume homogeneous and single granularity network i.e. all the links carry same number of wavelengths and all wavelengths support same bit rate. The parameter used for performance evaluation is blocking rate which is defined as the ratio of number of blocked lightpath requests over the total number of requested lightpath. The performance of proposed algorithm is evaluated, tested and validated for certain set of path set up demands. In order to implement the algorithm, a network of nine nodes is used. The designed algorithm implements sequential approach for path computation. Three different variations for path selection are designed namely First Fit (FF), Best Fit (BF) and Shortest Path (SP). The validation of algorithm is done by testing the performance of algorithm over Internet2 network and European Optical Network (EON) and comparing with existing techniques. The results obtained by presented algorithm are compared. It has been found that the response of presented algorithm is better than Simplified Lightpath Establishment and Regenerator Placement (sLERPS) due to less blocking rate being offered and comparable with Shortest Path First (SPF) and Longest Path First (LPF) algorithm. All the algorithms (BF, FF and SP) outperform than LPF as LPF consider longest path first. The increased path length will increase the impairment level of Q-matrix, thereby increasing the blocking probability of network. Moreover a signal which transverse more distance in the network will influence more number of already established lightpaths, resulting in call drops. The number of paths acting as potential candidates for lightpath setup through routing matrix plays an important role. The SPF algorithms use 10-shortest path search for lightpath set up, while the designed algorithm has no limit on number of potential paths for lightpath setup. This help in finding more optimum paths for specified connection request. Moreover, the performance of network changes with the network topology, number of nodes involved. The algorithm is implemented on a portion of EON network (only 9 nodes), while referred author used 15 nodes. The blocking probability for EON is high because the presented algorithm uses few nodes of EON (9 out of 15) so the number of available paths between source to destination is less, making it difficult to find path for high traffic load. The increased number of nodes might have increased the blocking rate of the network. The reason for difference in behavior of FF, BF and SP is same as already stated. The performance of SP is most reliable than all other algorithms. To further improve the system performance it is mandatory to improve the quality of link for signal transfer. The optimum placement of regenerators is one of the solutions for the same, where signal regeneration may takes place at certain pre-specified nodes. The existing strategies for regenerator placement are broadly classified as sparse regenerator placement and generation of transparent islands. [90] The sparse placement of regenerators seems to be a more economical solution since it distributes the regenerators throughout the network and probability of signal successfully reaching at destination node is more as it helps the signal maintaining its quality requirements. The nodes where regenerator must be placed may be calculated either by using an approximation algorithm or a heuristic algorithm where the incremental traffic and static traffic scenario are considered. In the present work, sparse placement of regenerators is considered for dynamic traffic scenario. Three different algorithms are presented using information of network topology, history of traffic flow in the network and presence of PLIs. The advantage of presented algorithms includes the simplicity in implementation and better network performance for the load distribution used during the optimization process. The algorithms include Maximum Utilized Node First, Traffic based regenerator placement and PLI constraint based regenerator placement. Three different cases of regenerator placement are considered i.e. when no regenerator is placed at all, when 40% of the available nodes are equipped with regeneration capability and when all the nodes have regeneration capability. The blocking rate of all the three cases is studied for a link supporting maximum of 15 wavelengths. Among the different routing algorithms SP algorithm responds best to all the three regenerator based strategies. Among the different regenerator based strategies, PLI based regenerator placement has maximum effect on reducing the blocking rate of the network. The performance of regenerator placement algorithms is compared with existing techniques. Although the performance of algorithm is comparable with existing results, but there is great scope in improvement due to numerous assumptions and network states considered by the author. The foremost reason is that referred work put limitation on the maximum number of wavelengths being transferred through the fiber. The number of impairments considered also influences the signal quality and overall call acceptance probability. The more complex, multigranuality networks may be explored in the future for further improvement in the field.