path protection in impairment-aware wdm networks

Path Protection in Impairment-Aware WDM

Networks

Ripudamanlall RamlallUniversity of Windsor, Windsor, Ontario, N9B 3P4, Canada

e-mail:[email protected]

April 6, 2013

Abstract

One of the major problems facing e-commerce, e-government and Internetusers, in general, is the reliability of data transfers across the network.The installation of optical networks has certainly helped to increase thespeed and reliability of data communication but network survivability stillremains a challenge. Imagine what would happen in the banking sector iftransactions cannot be processed because of poor network performance.The cost of providing separate systems with massive redundancies hasbeen excessively high. As such, a number of researchers have tried tosolve the problem of network reliability and, in particular, how to improvenetwork performance in impairment-aware WDM networks. This surveyis about their research.

1

Path protection in impairment-aware WDM networks 2

Contents

1 INTRODUCTION 31.1 The scope of the survey . . . . . . . . . . . . . . . . . . . . . . . 31.2 Why the survey is important . . . . . . . . . . . . . . . . . . . . 31.3 The resources used to identify related research . . . . . . . . . . 31.4 Research papers identified . . . . . . . . . . . . . . . . . . . . . . 31.5 Papers that were used as the basis of the survey . . . . . . . . . 31.6 The structure of the survey . . . . . . . . . . . . . . . . . . . . . 41.7 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 METHODS WHICH USE TRAFFIC GROOMING AND RE-GENERATOR PLACEMENT 52.1 Traffic Grooming Under Shared Path Protection . . . . . . . . . 52.2 Traffic Grooming with Dedicated Path Protection . . . . . . . . . 72.3 Traffic Grooming with Connection-level Protection . . . . . . . . 82.4 Section summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3 METHODS WHICH USE CONVENTIONAL RWA ALGO-RITHMS 93.1 RWA using lit or dark backup path protection . . . . . . . . . . . 93.2 RWA favouring resource sharing . . . . . . . . . . . . . . . . . . 113.3 RWA using a shared protection scheme . . . . . . . . . . . . . . . 123.4 RWA using a multicost approach . . . . . . . . . . . . . . . . . . 133.5 RWA using a modular approach . . . . . . . . . . . . . . . . . . . 143.6 RWA using a pre-processing module to order traffic demands . . 163.7 Section summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

4 METHODS WHICH USE SRLG APPROACH 174.1 Impairment-aware optimal diverse routing . . . . . . . . . . . . . 174.2 Performance enhancement for SRLG failures . . . . . . . . . . . 194.3 Section summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

5 CONCLUDING COMMENTS 20


1 INTRODUCTION

1.1 The scope of the survey

This survey concerns research which attempts to solve the problem of howto improve network performance. In particular, how to improve reliability inimpairment-aware WDM networks.

1.2 Why the survey is important

If the problem of improved network performance is solved, it would enable moreefficient and reliable transmission of data over optical networks.

1.3 The resources used to identify related research

Relevant research papers were found by searching Google Scholar with the key-words ”path protection”, ”protected path” OR ”protected-path”, ”impairmentaware” OR impairment-aware, and WDM. Various keywords and author nameswere also used to search the ACM Digital Library, IEEE Xplore and LectureNotes in Computer Science, through the University of Windsor Leddy Library.

It should be noted that some of the papers were not caught in the above-mentioned dragnets and were only found as a result of searching for the refer-ences given by selected papers.

1.4 Research papers identified

Thirteen (13) journal papers and sixteen (16) conference papers which areclosely related to the survey were identified. They are listed in the bibliog-raphy.

1.5 Papers that were used as the basis of the survey

Eleven (11) papers were chosen as the basis of this survey. There are six (6)journal papers and five (5) conference papers. The reasons for choosing themare as follows:

[Yang et al, April 2005] was chosen because it is a journal paper that is ex-actly on topic and was cited on three (3) of the list of eleven papers. It is alsoamong the first set of works published on this topic.

[Zhai et al, June 2007] was chosen because it is exactly on topic conferencepaper that was cited on three (3) of the list of eleven papers.

[Wang and Li, July 2007] was chosen because it is an exactly on topic jour-nal paper and was cited on one (1) of the list of eleven papers.


[Markidis and Tzanakaki, 2008] was chosen because it is an exactly on topicconference paper and was cited on one (1) of the list of eleven papers.

[dos Santos Rosa et al, January 2010] was chosen because it is an exactly ontopic journal paper that is written by a group of authors that published fourexactly on topic papers - dos Santos Rosa et al [2009] (two papers) and dosSantos Rosa et al [2010] (two papers).

[Shao et al, January 2010] was chosen because it is an exactly on topic journalpaper that was part of an important set of independent work published in 2010.

[Kokkinos et al, May 2010] was chosen because it is an exactly on topic confer-ence paper that was part of an important set of independent work published in2010.

[Georgakilas et al, August 2010] was chosen because it is an exactly on topicconference paper. Also, these authors are responsible for publishing three ex-actly on topic papers - Tzanakaki et al [2009] and Georgakilas et al [2010] (twopapers).

[Patel et al, September 2010] was chosen because it is an exactly on topic con-ference paper that was cited on one (1) of the list of eleven papers. Also, thisset of authors is part of a group of authors [Patel et al and Gao, et al] thatcollectively wrote six exactly on topic papers - Patel et al [2010] (three papers),Gao et al [2011] (two papers) and Gao et al [2012].

[Azodolmolky et al, October 2010] was chosen because it is an exactly on topicjournal paper that was part of an important set of independent work publishedin 2010.

[Gao et al, 2012] was chosen because it is one of the latest journal parers thatwas published that is exactly on topic. Also, this set of authors formed theother part of the group of authors referred to above - Patel et al [2010] (threepapers), Gao et al [2011] (two papers) and Gao et al [2012].

These papers are grouped by the general techniques used. Most of the re-searchers use the conventional RWA algorithm approach, but some also includethe placement of traffic grooming equipment and/or regenerators while othersinclude the shared risk link group (SRLG) approach.

1.6 The structure of the survey

The remainder of this survey is structured as follows:

Section 2 contains reviews of three (3) papers - Yang et al [2005], Patel etal [2010] and Gao et al [2012]. These papers are all concerned with the place-


ment of traffic grooming equipment and/or regenerators.

Section 3 contains reviews of six (6) papers - Zhai et al [2007], Markidis andTzanakaki [2008], dos Santos Rosa [2010], Kokkinos et al [2010], Georgakilas etal [2010] and Azodolmolky et al [2010]. These papers are all concerned withconventional RWA algorithm approach.

Section 4 contains reviews of two (2) papers - Wang and Li [2007] and Shaoet al [2010]. These two papers are both concerned with a method which usesthe SRLG approach.

Section 5 contains the concluding comments and Section 6 the list of references.

1.7 Observations

It was observed that a significant number of the papers were written by a fewauthors. dos Santos Rosa et al wrote four (4) of the papers while a group ofauthors working with Gao and Patel wrote six (6) of the papers. Also, fromtheir claims, it would appear that the authors were not aware of each others’work. It is thus, not surprising that sixteen (16) of the papers were publishedin 2009-2010, ten (10) alone in 2010.

2 METHODS WHICH USE TRAFFIC GROOM-ING AND REGENERATOR PLACEMENT

In this section, we review three papers describing research that attempts tosolve the problem by methods which use traffic grooming equipment and/orregenerator. The methods are traffic grooming and/or regenerator placementunder shared path protection, traffic grooming and/or regenerator placementwith dedicated path protection and traffic grooming and/or regenerator place-ment with connection-level protection.

2.1 Traffic Grooming Under Shared Path Protection

The problem addressedYang et al [2005] address the problem of protecting impairment-aware WDMnetworks against failures by assigning wavelengths and optical-electrical-optical(O/E/O) modules along the working and protection paths.

Previous work referred toThe authors do not refer to any previous work that attempts to solve the prob-lem.

The new algorithmThe researchers developed three approaches to solve the problem. An integer


linear program (ILP) approach that formulates the problem into a single IPLproblem, and two heuristic approaches (LOH and TSH) that first find an ini-tial solution by employing the divide-and-conquer and greedy principles. Adirected graph is used to represent the network, with each vertex representing anetwork node and each edge representing a fiber link that has a fixed number ofwavelengths. The objective is to minimize the number of O/E/O modules andwavelength links consumed by all the connection requests. In this approach k-shortest paths are chosen as candidates for the working path of each connection.Given the working path, another set of k-shortest paths is chosen as candidatesfor the protection path. A greedy wavelength and O/E/O assignment algorithmis then used to assign wavelengths and O/E/O modules along the working andprotection paths.

The LOH (local optimization heuristic) approach is used to improve the ini-tial solution, based on a reconfiguration evaluation procedure while the TSH(tabu-search heuristic), which the authors claim further optimize the solution,is based on a meta-heuristic for solving hard combinatorial optimization prob-lems.

ExperimentsExperiments were conducted on two networks; NET-A, a small-size networkwith 6 nodes and 9 bidirectional links, and NET-B, a network with 24 nodesand 43 bidirectional links. The ILP solution was tested only on NET-A, whilethe LOH and TSH were tested on both networks.

ResultsThe authors claim that their simulation results on NET-A revealed that theIPL solution becomes unacceptable when the number of wavelengths increasedto four but both the LOH and TSH found the optimal solution in all the sixcases in 60 seconds or less.

The cases are shown in the Table 1 below:

Case # Wavelengths Requests Running Time (s)1 2 3 5092 2 4 26223 3 4 32874 3 5 48265 4 5 7200*6 4 6 ?

Table1: Running times (in seconds) for solving the IPL problem in NET-A. Survivable

lightpath provisioning in WDM mesh networks under shared path protection and sign

al quality constraints. Yang et al [2005], page 1564.

The authors also claim that, in all cases, the solutions obtained by TSH arebetter than those obtained by LOH. However, both TSH and LOH show signif-


icant improvement over the initial solution, and TSH has a 4 percent improve-ment over LOH in terms of the number of O/E/O modules used and 5 percentimprovement over LOH in terms of the number of wavelength links used. In ad-dition, they claim that based on their observations, TSH generates improvementmove quickly than LOH and yields a better solution within the same amount oftime.

Claims madeThe authors claim that their results show that the tabu-search heuristic (TSH)outperforms the local optimization heuristic (LOH) and that both heuristic ap-proaches can handle large-scale problems with a modest time complexity.

Cited byThis paper was cited by Zhai et al [2007], dos Santos Rosa et al [2010] andKokkinos et al [2010].

2.2 Traffic Grooming with Dedicated Path Protection

The problem addressedPatel et al [2010] address the problem of protection of impairment-aware traffic-grooming WDM networks against single failures by focusing on efficient place-ment of equipment.

Previous work referred toThe authors did not refer to any previous work that attempt to solve the prob-lem.

The new algorithm and architectureThe authors developed a ROADM architecture that places traffic groomingequipment or a regenerator at each node. They also developed a heuristicalgorithm that is based on an auxiliary graph that consists of physical links(physical network topology), auxiliary links (routes that satisfied the impair-ments constraints) and virtual links (established lightpaths with sufficient sparecapacity for the request).

A new backup lightpath is established by finding the shortest path on the auxil-iary graph that is a link disjoint from the working path. Regenerators are thenplaced at all intermediate nodes along the backup path. Also, if there is sparecapacity, the algorithm establishes a separate working lightpath and a separatedisjoint backup lightpath.

ExperimentsExperiments were conducted, via simulations, on a 100-node meshed networktopology in which each link is assumed to have a bidirectional fiber. Line ratesof 10 Gbps, 40 Gbps and 100 Gbps with different optical reach were used in thesimulations.


ResultsThe authors claim that their results show that the cost of the grooming-onlyapproach is less than the cost of the regenerator-only approach and the cost ofthe combined traffic grooming and regenerator placement is always lower. Theyalso claim that based on the results: at low load, the 10 Gbps network has thelowest cost; at moderate load, the 40 Gbps network has the lowest cost; and athigh load, the 100 Gbps with reachability of 8 hops has the lowest cost.

Claims madeThe authors claim that the efficient placement of traffic grooming and regen-erators together for establishing survivable lightpaths reduces cost. They alsoclaim that as the traffic load increases, the network can further be improved byincreasing the line rate.

Cited byThis paper was cited by Gao et al [2012].

2.3 Traffic Grooming with Connection-level Protection

The problem addressedThe problem identified by Gao et al [2012] is protection of impairment-awareoptical WDM networks against failures by focusing on improving utilization ofnetwork resources.

Previous work referred toIn 2010, a subset of the authors proposed a combined traffic grooming and re-generation placement approach to support all traffic demands with minimumcost. [Patel et al 2010].

Shortcomings of previous work identifiedThe authors claim that the solution of Patel et al [2010] may not be cost orresource efficient for cases in which the lightpath is not fully occupied.

The new architecture and algorithmsThe authors developed a reconfigurable optical add-drop multiplexer (ROADM)node architecture. They state that this architecture has an all-optical wave-length switch fabric coupled with an electrical backbone, which together con-nects the various grooming and regenerator equipment.

The authors present two algorithms (a dedicated connection-level protectionalgorithm and a shared connection-level protection algorithm) that use auxil-iary graphs that take into consideration the weight and residual capacity of eachtype of link. They note that, in each algorithm, the auxiliary graph is first gen-erated and used to select the solution that generates the minimum network cost.


ExperimentsThe authors conducted experiments for both algorithms, on two simulated net-works. A randomly generated 100-node network topology, whose average nodedegree is 4, and a 14-node NFS network with a node degree of 2.7.

ResultsThe authors claim that their results show that when comparing connection-levelprotection with lightpath-level protection, among other parameters, connection-level protection achieves lower cost, but has a higher signaling cost when a light-path failure occurs.

Claims madeThe authors claim that their stimulation results show that their approach out-performs a lightpath-level protection approach, in terms of network cost. Theyalso claim to show that a network with higher line rate and longer optical reachcan reduce the network cost, when the traffic load increases.

2.4 Section summary

A summary of the three papers in this section is shown in the table below.

Year Title Authors Major contribution

2005 Survivable lightpath provisioning Yang, X., Equipment placementin WDM mesh networks under Shen L., and and shared pathshared path protection and Ramamurthy, B. protectionsignal quality constraints

2010 Survivable impairment-aware Patel, A., Jue, J., Equipment placementtraffic grooming with Wang, X., Zhang, Q., and dedicated pathdedicated path protection Palacharla, P., protectiondedicated path protection and Naito, T.

2012 Survivable impairment-aware Gao, C., Cankaya, H., Equipment placementtraffic grooming and regenerator Patel, A., Jue, J., and connection-levelplacement with connection-level Wang, X., Zhang, Q., protectionprotection Palacharla, P.,

and Sekiya, M.

3 METHODS WHICH USE CONVENTIONALRWA ALGORITHMS

In this section, we review six papers that attempt to solve the problem bymethods which use conventional routing and wavelength algorithms (RWA).

3.1 RWA using lit or dark backup path protection

The problem addressedZhai et al [2007] address the problem of protecting impairment-aware DWDM


networks against single link failures by considering dedicated path protectionschemes with lit and dark backup lightpaths.

Previous work referred toZhai et al [2007] refer to the work of Yang et al [2005].

Shortcomings of previous work identifiedThe authors claim that the solution of Yang et al [2005] did not consider fullytransparent networks.

The new architectureThe researchers developed a model that considered circuit-switched all opticalnetworks with no wavelength conversion. This model is based on two RWA al-gorithms Shortest Path with wavelength assignment (SP) and Highest Q-factor(HQ). In the HQ algorithm a shortest path algorithm is run on each wavelengthto find a candidate path on a specific wavelength. The end-to-end Q-factor isthen calculated and the path with the highest Q-factor is chosen for request.

For each traffic request, two new circuits (lightpaths) are tentatively estab-lished using one of the algorithms a primary circuit and a back circuit. In thebackup circuit, the protected path is either lit or kept dark (unlit), resultingin two dedicated path protection schemes. With the lit back path protectionscheme, SP runs a shortest path algorithm twice to compute two link-disjointpaths. Both paths are lit-up and the Quality of Transmission (QoT) blockingverification starts. With the dark backup path protection scheme, since onlythe primary path is lit, the QoT verification differs.

ExperimentsExperiments were conducted to study the network performance of the two dedi-cated path protection schemes using a down scaled version of the NFS topology.In this network 14 nodes and 21 bidirectional links were used. The blockingprobability and vulnerability ratio were used as the metric for performancecomparison.

ResultsThe authors claim that their simulation results revealed that the HQ algorithmwith lit backup paths performs better than the SP algorithm with dark backuppaths. They also claim that HQ improves vulnerability ratio over SP by a largemargin for the lit backup scheme than for the dark backup scheme.

For the dark backup scheme the authors claim that wavelength blocking domi-nates QoT blocking while in the lit backup scheme overall blocking probabilitieswere considerably higher; QoT blocking dominates wavelength blocking for theSP algorithm; and HP outperforms SP for blocking probability.


Claims madeThe authors claim that their results show that HQ outperforms SP in termsof blocking probability and vulnerability ratio in a certain traffic load range inboth dark and lit backup protection schemes.

Cited byThis paper was cited by dos Santos Rosa et al [2010], Shao et al [2010] andKokkinos et al [2010].

3.2 RWA favouring resource sharing

The problem addressedMarkidis and Tzanakaki [2008] address the problem of protecting impairment-aware transparent optical WDM networks against failures by focusing on light-path provisioning taking into consideration the optical signal performance degra-dation triggered by the physical impairments.


The new algorithmThe researchers used two steps to solve the RWA problems. Routing is im-plemented based on Dijkstras algorithm, to compute a primary and a backuppath. The Wavelength Assignment algorithms assign wavelength to primary andbackup paths, favouring resource sharing between the current demand and thealready established requests. After weights are assigned to the network links,Dijkstras algorithm is used to calculate the shortest path to find the primarypath. If no suitable path (based on BER) is found, the connection is blocked.After selecting the primary path, the process is repeated to select the backuppath, but the primary path is removed from the process. Random pick, first-fitor last-fit schemes are used to select a path from the list of candidate paths.

ExperimentsExperiments were conducted and results generated, by simulations, on the Pan-European test network defined by COST 239 that comprises of 11 nodes and26 bidirectional links each of capacity of 16 wavelengths. The behavior of last-fit, random-pick and first-fit for backup path were investigated. Physical layerconstraints were considered separately.

ResultsThe authors claim that their results show that last-fit wavelength assignmentscheme provides improved network performance of around 2 percent when com-pared to first-fit and significantly outperforms random-pick, since it can offer ablocking improvement of 12 percent for a wide range of traffic conditions. Theauthors also claim that network performance is significantly improved when the


impairment-aware routing scheme is compared to the impairment-unaware rout-ing scheme, in terms of blocking probability.

Claims madeThe authors claim that overall network performance can be improved whenwavelength assignment algorithms, that are able to offer high spare capacityutilization, are implemented. In addition, they claim that the proposed IARalgorithm incorporated in the provisioning of protected lightpaths is able toprovide guaranteed quality connections in a highly efficient manner.

Cited byThis paper was cited by Georgakilas et al [2010].

3.3 RWA using a shared protection scheme

The problem addresseddos Santos Rosa et al [2010] address the problem of protecting impairment-awareWDM networks against failures by focusing on shared path protection schemesthat take into consideration various physical impairments during path selection.

Previous work referred toThe authors refer to the work of Yang et al [2005], Zhai et al [2007] and Wangand Li [2007]

Shortcomings of previous work identifiedThe authors claim that Polarization Mode Dispersion (PMP), Amplifier Spon-taneous Emission Noise (ASE) and cross-talk (CT) impairments were not con-sidered in the previous work by other researchers.

The new algorithmsThe researchers developed, what they claim to be, two novel algorithms us-ing a shared protection scheme and impairments (PMP, ASE and CT) for theestablishment of lightpaths. These algorithms are First-fit IASPP and Random-selection IASPP. First-fit takes a traffic request and executes a shortest pathalgorithm to find the primary path, in order of wavelength. Not considering theprimary path, it then executes the same algorithm to search for a backup path.If a path is not available the request is blocked.

In the random-selection algorithm, there is no order and the layers are vis-ited randomly. Random selection takes a request and executes the shortestpath algorithm based on a random selection of a wavelength. It uses the samemethodology for selecting the backup path, not considering the primary path.If a path is not available the request is blocked.

ExperimentsExperiments were conducted using two network topologies: the NSF with 16


nodes and 25 bidirectional links and the USA with 24 nodes and 43 bidirectionallinks. The performance metric used is the blocking probability.

ResultsThe authors claim that in the NSF network topology, the results of FFIASPPand RSIASSP when compared to FFSSP and RSSSP (without impairment-aware) show: a significant increase in blocking probability is notices for impairednetworks; the FFIASSP reduced the blocking probability from 0.32 to 0.2, whilethe RSIASPP reduced it from 0.3 to 0.17. In terms of vulnerability, the ratiosvary very little.

In the USA network topology, the blocking probability increased in both awareand unaware algorithms. In terms of vulnerability, the figures were lower thanthose obtained in the NSF topology.

Claims madeThe authors claim that results show that this approach decreases blocking, yetmaintains a similar level of vulnerability.

3.4 RWA using a multicost approach

The problem addressedKokkinos et al [2010] address the problem of protecting impairment-aware WDMoptical networks against failures by selecting the primary path and the backuppath, jointly.

Previous work referred toThe authors refer to the work of Yang et al [2005], Zhai et al [2007] and Azodol-molky et al [2009].

Shortcomings of previous work identifiedThe authors claim that in the work of Yang et al [2005] only considered two lin-ear constraints (PMD and ASE) and ignored all the other physical constraints.They also claim that in the work of Zhai et al [2007], keeping the backup pathlit worsen the impairments for other lightpaths and keeping the backup pathdark can lead to increased traffic restoration time. They further claim that thework of Azodolmolky et al [2009] was limited to transparent all optical networksand only considered dedicated path protection for offline demands.

The new idea and algorithmThe authors developed a multicost approach that they claim is significantly morepowerful than single cost routing. A vector of cost parameters, which includesthe utilization of wavelengths, is assign to each link. These cost parameters areused to calculate the cost vector of a path. In this multicost approach morethan one path is calculated between two nodes and, as such, minimization andmaximization can be used to arrive at the best path. This approach consists


of two phases. The first phase uses an algorithm that is a generalization of theDijkstras algorithm which obtains a set of candidate non-dominated lightpathsfor a given connection request. In the second phase, a primary lightpath is se-lected to serve the connection request while a second lightpath, from the sameset is selected as the backup path.

ExperimentsThe authors conducted simulation experiments considering an all-optical net-work where connections arrived dynamically and have to be served as they come.The blocking probability of several multicost based algorithms as a function ofthe number of available wavelengths, performing 1+1 and 1:1 protection weremeasured.

ResultsThe authors claim that the results show that in case of 1+1 protection, thesumQ and bestQ-bestQ policies result in similar blocking probability and in thecase of 1:1 protection, the differences between the different protection policiesare minimized. They also claim that there is a level of trade off between thelevel of protection (1+1, 1:1 and no protection). In addition, the authors claimthat the usage of the worse case interference assumption for the selection ofbackup lightpath reduces the blocking probability.

Claims madeThe authors claim that, based on their experimental observations, the joint se-lection of the primary and backup lightpaths is generally more preferable thanselecting them separately. They also claim to show that the selection of backuplightpath, based on the worse case interference assumption can be quite ben-eficial for the survivability of the backup paths, when the network is heavilyloaded. In addition, they claim that the problem is NP-complete.

3.5 RWA using a modular approach

The problem addressedGeorgakilas et al [2010] address the problem of protecting impairment-awareWDM networks against failures by focusing on performance evaluation in impairment-aware routing under single and double link failures in identifying the primaryand backup paths.

Previous work referred toThe authors refer to the work of Markadis and Tzanakaki [2008].

Shortcomings of previous work identifiedThe authors claim that the work of Markadis and Tzanakaki [2008] did not takeinto consideration double-link failures.


The new idea and algorithmThe authors developed a modular approach that divides the problem into sev-eral modules. The RWA problem is solved in two steps: routing and wavelengthassignment (RWA) for the primary path and RWA for the backup path. Bothsteps are implemented using Dijkstras algorithm. The primary path for eachrequest is computed by the impairment-aware RWA scheme and uses a first-fitalgorithm based on the quality of the paths (a predefined BER threshold) forwavelength assignment. The backup path is computed using IA and minimum-hop routing and uses the weighted last-fit algorithm for wavelength assignment(also based on the BER threshold)

ExperimentsFour Experiments, based on simulations, were evaluated using two networktopologies: the COST 239 topology and the NSFNET topology. The scenariosare shown in the table below:

Scenario Backup Path Routing Reinforced Sharing

Sc1 IAR Yes

Sc2 IAR No

Sc3 min-hop Yes

Sc4 min-hop NoTable 1: Evaluation Scenarios Definitions, Performance evaluation of impairment-

aware routing under single- and double-link failures. Georgakilas et al [2010], page

637.

ResultsThe authors claim that they observed that when IAR was applied for the backuppath computation, the total blocking probability for the COST 239 topology wasmuch lower compared with the min-hop; for all load conditions, the blockingprobability is reduced by more than 50 percent. In the NSFNET, the IAR alsodemonstrated reduction in blocking probability. They also claimed that whenreinforced sharing is applied there is better resource utilization.

Claims madeThe authors claim that when the quality of the formed paths is taking into ac-count in the process of routing, the physical performance of the selected pathsis shown to be better, compared with the traditional routing schemes that con-siders only bandwidth availability. They also claim that the impairment-awarerouting scheme provides better performance under single- and dual-links fail-ures. In addition, they claim that the consideration of impairments in therouting algorithm for both the primary and backup paths provides a reducednetwork blocking probability.


3.6 RWA using a pre-processing module to order trafficdemands

The problem addressedAzodolmolky et al [2010] address the problem of protecting impairment-awareWDM networks against failures by focusing on offline physical layer impairmentaware routing with dedicated path protection consideration.


The new algorithmsIn this performance evaluation, the authors developed three algorithms. Theyenhanced two existing IA-RWA algorithms (the RS-RWA algorithm and an ILP-based RWA algorithm) that did not considered dedicated path protection. Thetwo offline IA-RWA algorithms were extended to consider dedicated path pro-tection. The performances of these two algorithms were compared with that ofa new IA-RWA algorithm (the Rahyab algorithm). The new algorithm uses ademand pre-processing module to order the traffic demands and a process toselect a lightpath form the candidate lightpath list. Traffic demands can beprotected or unprotected.

ExperimentsExperiments were conducted based on simulations that used the Deutsche TelekomNational Network (DTNet) which 14 nodes and 23 bidirectional links, with anaverage node degree of 3.29.

ResultsThe authors claim that their results show that the performance of the ILP-RWA-LU (1 SP) is better than the original ILP-RWA algorithms (both 1 SPand 2 SP). They also claim that their results indicate that increasing the num-ber of candidate shortest paths also increases the blocking rate of ILP-RWA andILP-RWA-LU algorithms.

Claims madeThe authors claim that their enhancements to the RS-RWA heuristic (that is,RS-RWA-Q) reduced the blocking rate of demands by an average of 35 percentfrom different loads compared to the original scheme (RS-RWA); the perfor-mance of ILP-RWA-LU (1 sp) formulation (enhanced IRL-RWA) is also im-proved by 71 percent; and the Rehab algorithm performs better than the twoalgorithms with respect to the blocking rate performance metric.


3.7 Section summary

A summary of the six papers in this section is shown in the table below.


2007 Performance of dedicated Zhai, Y., RWA and dedicatedpath protection in Pointurier, Y., path protectiontransmission-impaired Subramanian, S. andDWDM networks Brandt-Pearce, M.

2008 Routing and Wavelength Markidis, G. and RWA and lightpathassignment in survivable Tzanakaki, A. provisioningWDM networks underphysical layer constraints

2010 Path protection WDM dos Santos Rosa, S., RWA and sharedJan networks with Drummond, A. and path protection

impaired-transmission da Donseka, N.

2010 Path protection in Kokkinos, P., RWA and jointMay WDM networks with Manousakis, K. and selection of primary

quality of transmission Varvarigos, E. and backup pathslimitations

2010 Performance evaluation of Georgakilas, K., RWA and singleAug impaired-aware routing Katrinis, K., and double-link

under single- and Tzanakaki, A. and failuresdouble-link failures Madsen, O.

2010 A novel offline physical Azodolmolky, S., RWA and dedicatedOct layer impairments aware Klinkowski, M., path protection

RWA algorithm with Pointurier, Y.,dedicated path protection Angelou, M.,considerations Careglio, D.,

Sol-Pareta, J. andTomkos, I.

4 METHODS WHICH USE SRLG APPROACH

In this section, we review two papers that attempt to solve the problem by meth-ods which use shared risk link group (SRLG). The methods are impairment-aware optimal routing considering SRLG and performance enhancement forSRLG failures.

4.1 Impairment-aware optimal diverse routing

The problem addressedWang and Li [2007] address the problem of protecting impairment-aware WDMnetworks against failures by selecting reliable lightpaths to be used as primaryand backup paths.


Previous work referred toThe authors refer to the work of Zang et al [2002]

Shortcomings of previous work identifiedThe authors claim that in the work of Zang et al [2002], the searching processis not complete and the solution is not optimal, in terms of the total cost of thetwo SRLG-disjoint paths, for Separate Path Selection (SPS), Joint Path Selec-tion (JPS) and Low-cost Pair of SRLG-Disjoint (LPSD). They also claim thatthe Modified Weight (MW) gives optimal solution only when the two paths arerequired to be link or node disjoint but not SRLG-disjoint. More so, they fur-ther claim that the algorithm can only be applied in a static traffic environment.

The new algorithmsThe authors developed, what they claim to be, two novel algorithms to calculatean optimal path pair for a given connection requirement from a source node toa destination node (s-d pair). The optimal path has two definitions and thus,two versions of the problem are defined (IAOPP-I and IAOPP-II). The firstalgorithm has two versions (MUCOPPON-I and MUCOPPON-II) which solvethe IAOPP-I and IAOPP-II problems respectively. In MUCOPPON, the costof a path pair has two definitions. In MUCOPPON-I, it is defined as the sumof primary costs of the two paths while in MUCOPPON-II, it is defined as thelower primary cost of the two paths. The other algorithm (DEPPON) can beused to solve the IAOPP-II problem only. It tries to find the optimal path pairin a path-by-path manner.

ExperimentsExperiments were conducted to study the MUCOPPON and DEPPON algo-rithms in two network topologies: the Pan-European test network COST 239with 11 nodes and 26 links, and the ItalianNet with 21 nodes and 36 links. TheSuccess Ratio and average primary cost were used as the metric for performancecomparison. For comparison, another algorithm, the MSPS algorithm was used.

ResultsThe results are shown in Table 1 and Table 2 on the following page.

Claims madeThe authors claim that their simulation results prove the feasibility and goodperformance of MUCCOPON and DEPPON to address the special issues ofrouting lightpaths in wavelength-routed all-optical networks.

Cited byThis paper was cited by dos Santos Rosa et al [2010].


Table 1 and Table 2

COST239, NW = 32, DMax = 2, 500(B = 40, DPMD = 0.05), Hmax = 10, averageload =

100Erlang.MUCOPPON-I MUCOPPON-II DEPPON MSPS

SR 0.998 0.998 0.998 0.374

AvgPriCost Working path 100.101 99.193 99.193 65.578

Path pair 285.385 325.635 286.632 203.216Table1: Impairment aware optimal diverse routing for survivable optical networks.

Wang and Li [2007], page 149.

ItalianNet,NW = 32, DMax = 2, 500(B = 40, DPDM = 0.05), Hmax = 10, averageload =


SR 0.825 0.825 0.825 0.405


Path pair 620.589 659.981 636.908 412.58Table2: Impairment aware optimal diverse routing for survivable optical networks.

Wang and Li [2007], page 149.

4.2 Performance enhancement for SRLG failures

The problem addressedShao et al [2012] address the problem of protecting impairment-aware WDMnetworks against SRLG failures by selecting working and backup paths withSRLG-disjoint constraints.

Previous work referred toThe authors refer to the work of Zhai et al [2007]

Shortcomings of previous work identifiedThe authors claim that in the work of Zhai et al [2007] connections with a longerworking path or backup path will be rejected.

The new ideaThe new idea considers two policies: Policy I and Policy II. Policy I is used tosatisfy impairment constraints. It computes the two least impairment paths.The first path is computed the working path and then another path is computefor the backup path. Policy II is based on capacity efficiency. It tries to benefitfrom both the shortest path and the least impaired path by choosing them in-telligently.

ExperimentsThe authors conducted two experiments, based on simulations, one using a 14-node network topology and the other using a 28-node network topology. They


compare the Policy I and Policy II based physical impairment-aware shared-path protection (SPP) with impairment-unaware SPP that does not considerphysical impairment.

ResultsThe authors claim that the results show that the two impairment-aware methodssignificantly outperform impairment-unaware SSP with the presence of physicalimpairment constraints. Also, they claim that the results show that generally,the impairment-aware SPP based on Policy II outperforms that based on PolicyI in terms of blocking probability.

Claims madeThe authors claim that when compared with impairment-unaware SRLG fail-ure protection, impairment-aware performs much better in terms of blockingprobability especially with strong physical impairment constraints. They alsoclaim that impairment-aware SRLG failure protection can significantly reducephysical-layer blocking probability, and the algorithm based on Policy II achievesa good balance between capacity efficiency and physical impairment require-ments. In addition they claim that the problem is NP-complete.

4.3 Section summary

A summary of the papers in this section is shown in the table below.


2007 Impairment aware optimal Wang, S., Consideration fordiverse routing for and Li, L. protection andsurvivable optical networks SRLG failures

2010 Performance enhancement for Shao, X., Yeo, K., Consideration forimpairment-aware SRLG failure Zhou, L., Wang, Y., protection andprotection in wavelength-routed and Bai, Y. SRLG failuresoptical networks

5 CONCLUDING COMMENTS

Twenty nine (29) papers related to the topic of the survey were identified andeleven of them were reviewed in detail. In doing so, the following observationswere made:

1. There appears to be a limited set of researchers doing work in this area,as four (4) of the papers were written by members of one research group. Thisgroup consists of three members (dos Santos Rosa, S., Drummond, A., andda Fonseca, N.), all from the Institute of Computing, University of Campinas,Campinas, Brazil. Two of the papers were published in 2009 - dos Santos Rosaet al [2009] and another two in 2010 - dos Santos Rosa et al [2010].


Another six (6) of the papers were written by an extended group of researchers.This extended group consists of nine researchers (Gao, C., Cankaya, H., Pa-tel, A., Jue, J., Wang, X., Zhang, Q., Palacharla, P., Naito, T., andSekiya, M.). Three of the papers were published in 2010 - Patel et al [2010],two more in 2011 - Gao et al [2011] and yet another in 2012 - Gao et al [2012].

2. The authors were probably not aware of each other work, as many of thethem claimed to be the first to provide a particular methodology. This is prob-ably a result of the fact that they were working at the same time, and ten (10)of the papers were published in 2010.

To provide a better understanding of the nature of the research in this area,the table below list the eleven most important papers, which papers were re-ferred to (cites) and who cites the original paper (cited by).

Paper Cites Cited by

Yang et al 2005 Zhai et al 2007dos Santos Rosa et al 2010Kokkinos et al 2010

Zhai et al 2007 Yang et al 2005 dos Santos Rosa et al 2010Shao et al 2010Kokkinos et al 2010

Wang and Li 2007 Zang et al 2002 dos Santos Rosa et al 2010

Markidis and Georgakilas et al 2010Tanakaki 2008

dos Santos Rosa et al 2010 Yang et al 2005Zhai et al 2007Wang and Li 2007

Shao et al 2010 Zhai et al 2007

Kokkinos et al 2010 Yang et al 2005Zhai et al 2007Azodolmolky et al 2009

Georgakilas et al 2010 Markidis andTanakaki 2008

Patel et al 2010 Gao et al 2012

Azodolmolky et al 2010

Gao et al 2012 Patel et al 2010

The researchers mentioned the following future work:

a) Yang et al [April 2005] state that the more generic shared risk link group(SRLG)-diverse protection can be incorporated into the problem and can behandled by enhancing their solution approaches.

b) Zhai et al [June 2007] state that they intend to develop less time-consumingQoT-aware algorithm with blocking probability and vulnerability ratio perfor-mance comparable to those of the Highest Q factor algorithm. They also state


that more complicated protection schemes (such as shared protection) and fail-ure types (such as Shared Risk Link Group and multi-link failures) will need tobe investigated.

c) Shao et al [January 2010] state that future work includes studying the prob-lem of how to support quality of Protection (QoP) classes of surviving SRLGfailures and Quality of Transmission (QoT) by differentiated QoP and QoT pro-visioning.

d) Azodolmolky et al [October 2010] state that proper integration of IA-RWAalgorithms with control plane and its impact on control plane operation areamong their ongoing activities.


References

[1] Azodolmolky, S., Klinkowski, M., Pointurier, Y., Angelou, M.,Careglio, D., Sol-Pareta, J., and Tomkos, I. 2010. A novel offlinephysical layer impairments aware RWA algorithm with dedicated path pro-tection consideration. Journal of Lightwave Technology, 28, 20, 3029-3040.

[2] Azodolmolky, S., Pointurier, Y., Angelou, M., Sole Pareta, J.,and Tomkos, I. 2009. An offline impairment aware RWA algorithm withdedicated path protection consideration. Optical Fiber Communication -incudes post deadline papers, 2009. OFC 2009. Conference, 1-3.

[3] Bao, N.-H., Li, L.-M., Luo, H.-B., Yu, H.-F., and Zhang, Z.-Z. 2012. Impairment aware sharing constraint relaxed path protectionin translucent optical networks. Opt. Eng. 51(4), 045002 Apr 06, 2012.doi:10.1117/1.OE.51.4.045002.

[4] dos Santos Rosa, S., Drummond, A., and da Fonseca, N. 2009.Lightpath establishment in WDM networks with best effort shared pathprotection in impaired-transmissions. Communications, 2009. ICC ’09.IEEE International Conference, 1-6.

[5] dos Santos Rosa, S., Drummond, A., and da Fonseca, N. 2009.Performance of best effort shared path protection mechanism under physi-cal impairments in WDM networks. Optical Network Design and Modeling,2009. ONDM, International Conference, 1-6.

[6] dos Santos Rosa, S., Drummond, A., and da Fonseca, N. 2010.Path protection WDM networks with impaired-transmission. Photonic Net-work Communications, 19, 2 , 212-222.

[7] dos Santos Rosa, S., Drummond, A., and da Fonseca, N. 2010.Shared path protection with differentiated reliability in transmission im-paired WDM networks. Communications (ICC), 2010 IEEE InternationalConference, 1-6.

[8] Gao, C., Cankaya, H., Patel, A., Jue, J., Wang, X., Zhang, Q.,Palacharla, P., and Sekiya, M. 2011. Survivable impairment-awaretraffic grooming and regenerator placement with dedicated connection levelprotection. Communications (ICC), 2011 IEEE International Conference,1-5.

[9] Gao, C., Cankaya, H., Patel, A., Jue, J., Wang, X., Zhang, Q.,Palacharla, P., and Sekiya, M. 2011. Survivable impairment-awaretraffic grooming and regenerator placement with shared connection-levelprotection. Transparent Optical Networks (ICTON), 2011 13th Interna-tional Conference, 1-4.


[10] Gao, C., Cankaya, H., Patel, A., Jue, J., Wang, X., Zhang, Q.,Palacharla, P., and Sekiya, M. 2012. Survivable impairment-awaretraffic grooming and regenerator placement with connection-level protec-tion. Optical Communications and Networking, IEEE OSA Journal. 4, 3,259-270.

[11] Georgakilas, K., Katrinis, K., Tzanakaki, A., and Madsen, O.2010. Impact of dual-link failures on impairment-aware routed networks.Transparent Optical Networks (ICTON), 2010 12th International Confer-ence, 1-4.

[12] Georgakilas, K., Katrinis, K., Tzanakaki, A., and Madsen,O. 2010. Performance evaluation of impairment-aware routing undersingle- and double-link failures. Optical Communications and Networking,IEEE/OSA Journal 2, 8, 633-641.

[13] Guo, L., and Ji, Y. 2007. Shared protection schemes for multi-granularityoptical networks. Proc. SPIE 6784, Network Architectures, Management,and Applications V, 678410 November 19, 2007; doi:10.1117/12.745381

[14] Kokkinos, P., Manousakis, K., and Varvarigos, E. 2010. Path pro-tection in WDM networks with quality of transmission limitations. Com-munications (ICC), 2010 IEEE International Conference, 1-5.

[15] Lee, Y., and Mukherjee, B. 2004. Traffic engineering in next-generationoptical networks. Communications Surveys and Tutorials, IEEE 6, 3, 16-33.

[16] Markidis, G., and Tzanakaki, A. 2008. Routing and wavelength as-signment algorithms in survivable WDM networks under physical layerconstraints. Broadband Communications, Networks and Systems, 2008.BROADNETS 2008. 5th International Conference, 191-196.

[17] Patel, A., Gao, C., Jue, J., Wang, X., Zhang, Q., Palacharla,P., and Naito, T. 2010. Traffic grooming and regenerator placementin impairment-aware optical WDM networks. Optical Network Design andModeling (ONDM), 2010 14th Conference, 1-6.

[18] Patel, A., Jue, J., Wang, X., Zhang, Q., Palacharla, P., andNaito, T. 2010. Survivable impairment-aware traffic grooming with ded-icated path protection. Optical Communication (ECOC), 2010 36th Euro-pean Conference and Exhibition, 1-3.

[19] Patel, A., Jue, J., Wang, X., Zhang, Q., Palacharla, P., andNaito, T. 2010. Survivable traffic grooming with impairment constraints.Computer Communications and Networks (ICCCN), 2010 Proceedings of19th International Conference, 1-6.


[20] Saradhi, C., and Subramaniam, S. 2009. Physical layer impairmentaware routing (PLIAR) in WDM optical networks: issues and challenges.Communications Surveys and Tutorials, IEEE 11, 4, 109-130.

[21] Shao, X., Cheng, X., Yeo, Y. K., Zhou, L., Wang, Y., and Bai,Y. 2010. Performance enhancement for impairment-aware SRLG failureprotection in wavelength-routed optical networks. Photonic Network Com-munications 20, 1, 1-9.

[22] Shao, X., Zhou, L., Chai, T. Y., Saradhi, C. V., and Wang, Y.2008. Reducing vulnerability of shared-path protection subject to double-link failures in WDM mesh networks. Photonic Network Communications15, 2, 171-181

[23] Tomkos, I., Azodolmolky, S., Angelou, Marianaand Klonidis,D., Ye, Y., Saradhi, C. V., Salvadori, E., Zanardi, A., andPiesiewicz, R. 2008. Impairment aware networking and relevant resiliencyissues in all-optical networks. Optical Communication, 2008. ECOC 2008.34th European Conference, 1-4.

[24] Tzanakaki, A., Georgakilas, K., Katrinis, K., Wosinska, L., Ji-rattigalachote, A., and Monti, P. 2009. Network performance im-provement in survivable WDM networks considering physical layer con-straints. Transparent Optical Networks, 2009. ICTON ’09. 11th Interna-tional Conference, 1-4

[25] Wang, S., and Li, L. 2007. Impairment aware optimal diverse routingfor survivable optical networks. Photonic Network Communications April2007, 13, 2, 139-154.

[26] Wosinska, L., Jirattigalachote, A., Monti, P., Tzanakaki, A.,and Katrinis, K. 2009. Lightpath routing considering differentiated phys-ical layer constraints in transparent WDM networks. Communications andPhotonics Conference and Exhibition (ACP), 2009 Asia, 1-8.

[27] Yang, X., Shen, L., and Ramamurthy, B. 2005. Survivable lightpathprovisioning in WDM mesh networks under shared path protection andsignal quality constraints. Lightwave Technology, Journal of April 2005.23, 4, 1556-1567.

[28] Zang, H., Ou, C., and Mukherjee, B. 2002. Path-protection Routingand wavelength assignment in WDM mesh networks under shared-risk-group constraints. Proc. SPIE APOC 2002, Beijing, China, Oct. 2002,4585, 49-59.

[29] Zhai, Y., Pointurier, Y., Subramanian, S., and Brandt-Pearce,M. 2007. Performance of dedicated path protection in transmission-impaired DWDM networks. In: Proceedings of IEEE International con-ference on communication, 2342-2347.


Annotations

Yang, X., Shen, L., and Ramamurthy, B. 2005. Survivablelightpath provisioning in WDM mesh networks under shared pathprotection and signal quality constraints. Lightwave Technology,Journal of April 2005. 23, 4, 1556-1567.

The problem which the researcher/authors addressedThe authors address the problem of protecting impairment-awareWDM networks against failures by assigning wavelengths and optical-electrical-optical (O/E/O) modules along the working and protec-tion paths.

Previous work by other authors referred to by the authorsThe authors do not refer to any previous work that attempts tosolve the problem.

The new idea, algorithm, architecture, protocol, etc.The researchers developed three approaches to solve the problem.An integer linear program (ILP) approach that formulates the prob-lem into a single IPL problem, and two heuristic solutions approaches(LOH and TSH) that first find an initial solution by employing thedivide-and-conquer and greedy principles. A directed graph is usedto represent the network, with each vertex representing a networknode and each edge representing a fiber link that has a fixed numberof wavelengths. The objective is to minimize the number of O/E/Omodules and wavelength links consumed by all the connection re-quests. In this approach k-shortest paths are chosen as candidatesfor the working path of each connection. Given the working path,another set of k-shortest paths is chosen as candidates for the protec-tion path. A greedy wavelength and O/E/O assignment algorithmis then used to assign wavelengths and O/E/O modules along theworking and protection paths.

The LOH (local optimization heuristic) approach is used to improvethe initial solution, based on a reconfiguration evaluation procedurewhile the TSH (tabu-search heuristic), which the authors claim fur-ther optimize the solution, is based on a meta-heuristic for solvinghard combinatorial optimization problems.


Experiments and/or analyses conductedExperiments were conducted on two networks; NET-A, a small-sizenetwork with 6 nodes and 9 bidirectional links, and NET-B, a net-work with 24 nodes and 43 bidirectional links. The ILP solutionwas tested only on NET-A, while the LOH and TSH were tested onboth networks.

Results that the authors claimed to have achievedThe authors claim that their simulation results on NET-A revealedthat the IPL solution becomes unacceptable when the number ofwavelengths increased to four but both the LOH and TSH foundthe optimal solution in all the six cases in 60 seconds or less.

The cases are shown in the Table 1 below:

Case # Wavelengths Requests Running Time (s)1 2 3 5092 2 4 26223 3 4 32874 3 5 48265 4 5 7200*6 4 6 ?

Table1: Running times (in seconds) for solving the IPL problem in NET-A. Survivable

lightpath provisioning in WDM mesh networks under shared path protection and sign

al quality constraints [2005]. Yang et al, page 1564.

The authors also claim that, in all cases, the solutions obtained byTSH are better than those obtained by LOH. However, both TSHand LOH show significant improvement over the initial solution, andTSH has a 4 percent improvement over LOH in terms of the num-ber of O/E/O modules used and 5 percent improvement over LOHin terms of the number of wavelength links used. In addition, theyclaim that based on their observations, TSH generates improvementmove quickly than LOH and yields a better solution within the sameamount of time.

Claims made by the authorsThe authors claim that their results show that the tabu-search


heuristic (TSH) outperforms the local optimization heuristic (LOH)and that both heuristic approaches can handle large-scale problemswith a modest time complexity.

Zhai, Y., Pointurier, Y., Subramanian, S., and Brandt-Pearce, M. 2007. Performance of dedicated path protection intransmission-impaired DWDM networks. In: Proceedings of IEEEInternational conference on communication, 2342-2347.

The problem which the researcher/authors addressedThe authors address the problem of protecting impairment-awareDWDM networks against single link failures by considering dedi-cated path protection schemes with lit and dark backup lightpaths.

Previous work by other authors referred to by the authors[Yang et al 2005]

Shortcomings of previous workThe authors claim that the solution of Yang et al [2005] did notconsider fully transparent networks.

The new idea, algorithm, architecture, protocol, etc.The researchers developed a model that considered circuit-switchedall optical networks with no wavelength conversion. This model isbased on two RWA algorithms Shortest Path with wavelength as-signment (SP) and Highest Q-factor (HQ). In the HQ algorithm ashortest path algorithm is run on each wavelength to find a can-didate path on a specific wavelength. The end-to-end Q-factor isthen calculated and the path with the highest Q-factor is chosen forrequest.

For each traffic request, two new circuits (lightpaths) are tenta-tively established using one of the algorithms a primary circuit anda back circuit. In the backup circuit, the protected path is eitherlit or kept dark (unlit), resulting in two dedicated path protectionschemes. With the lit back path protection scheme, SP runs a short-est path algorithm twice to compute two link-disjoint paths. Bothpaths are lit-up and the Quality of Transmission (QoT) blocking


verification starts. With the dark backup path protection scheme,since only the primary path is lit, the QoT verification differs.

Experiments and/or analyses conductedExperiments were conducted to study the network performance ofthe two dedicated path protection schemes using a down scaled ver-sion of the NFS topology. In this network 14 nodes and 21 bidirec-tional links were used. The blocking probability and vulnerabilityratio were used as the metric for performance comparison.

Results that the authors claimed to have achievedThe authors claim that their simulation results revealed that the HQalgorithm with lit backup paths performs better than the SP algo-rithm with dark backup paths. They also claim that HQ improvesvulnerability ratio over SP by a large margin for the lit backupscheme than for the dark backup scheme.

For the dark backup scheme the authors claim that wavelengthblocking dominates QoT blocking while in the lit backup schemeoverall blocking probabilities were considerably higher; QoT block-ing dominates wavelength blocking for the SP algorithm; and HPoutperforms SP for blocking probability

Claims made by the authorsThe authors claim that their results show that HQ outperforms SPin terms of blocking probability and vulnerability ratio in a certaintraffic load range in both dark and lit backup protection schemes.

Wang, S., and Li, L. 2007. Impairment aware optimal diverserouting for survivable optical networks. Photonic Network Commu-nications April 2007, 13, 2, 139-154.

The problem which the researcher/authors addressedThe authors address the problem of protecting impairment-awareWDM networks against failures by selecting reliable lightpaths tobe used as primary and backup paths.

Previous work by other authors referred to by the authors


[Zang et al 2002]

Shortcomings of previous workThe authors claim that in the work of Zang et al [2002], the search-ing process is not complete and the solution is not optimal, in termsof the total cost of the two SRLG-disjoint paths, for Separate PathSelection (SPS), Joint Path Selection (JPS) and Low-cost Pair ofSRLG-Disjoint (LPSD). They also claim that the Modified Weight(MW) gives optimal solution only when the two paths are requiredto be link or node disjoint but not SRLG-disjoint. More so, theyfurther claim that the algorithm can only apply in static traffic en-vironment.

The new idea, algorithm, architecture, protocol, etc.The authors developed, what they claim to be, two novel algorithmsto calculate optimal path pair for a given connection requirementfrom a source node to a destination node (s-d pair). The opti-mal path has two definitions and thus, two versions of the prob-lem are defined (IAOPP-I and IAOPP-II). The first algorithm hastwo versions (MUCOPPON-I and MUCOPPON-II) which solve theIAOPP-I and IAOPP-II problems respectively. In MUCOPPON,the cost of a path pair has two definitions. In MUCOPPON-I, itis defined as the sum of primary costs of the two paths while inMUCOPPON-II, it is defined as the lower primary cost of the twopaths. The other algorithm (DEPPON) can be used to solve theIAOPP-II problem only. It tries to find the optimal path pair in apath-by-path manner.

Experiments and/or analyses conductedExperiments were conducted to study the MUCOPPON and DEP-PON algorithms in two network topologies: the Pan-European testnetwork COST 239 with 11 nodes and 26 links, and the ItalianNetwith 21 nodes and 36 links. The Success Ratio and average primarycost were used as the metric for performance comparison. For com-parison, another algorithm, the MSPS algorithm was used.

Results that the authors claimed to have achievedThe results are shown in Table 1 and Table 2 below:


COST239, NW = 32, DMax = 2, 500(B = 40, DPMD = 0.05), Hmax = 10, averageload =

100Erlang.

MUCOPPON-I MUCOPPON-II DEPPON MSPS

SR 0.998 0.998 0.998 0.374


Path pair 285.385 325.635 286.632 203.216Table1: Impairment aware optimal diverse routing for survivable optical networks

[2007] Wang, S and Li, L, page 149.

ItalianNet,NW = 32, DMax = 2, 500(B = 40, DPDM = 0.05), Hmax = 10, averageload =


SR 0.825 0.825 0.825 0.405


Path pair 620.589 659.981 636.908 412.58Table2: Impairment aware optimal diverse routing for survivable optical networks

[2007] Wang, S and Li, L, page 149.

Claims made by the authorsThe authors claim that their simulation results prove the feasibil-ity and good performance of MUCCOPON and DEPPON to ad-dress the special issues of routing lightpaths in wavelength-routedall-optical networks.

Markidis, G., and Tzanakaki, A. 2008. Routing and wave-length assignment algorithms in survivable WDM networks underphysical layer constraints. Broadband Communications, Networksand Systems, 2008. BROADNETS 2008. 5th International Confer-ence, 191-196.

The problem which the researcher/authors addressedThe authors address the problem of protecting impairment-awaretransparent optical WDM networks against failures by focusing onlightpath provisioning taking into consideration the optical signalperformance degradation triggered by the physical impairments.

Previous work by other authors referred to by the authorsThe authors do not refer to any previous work that attempts to


solve the problem.

The new idea, algorithm, architecture, protocol, etc.The researchers used two steps to solve the RWA problems. Routingis implemented based on Dijkstras algorithm, to compute a primaryand a backup path. The Wavelength Assignment algorithms assignwavelength to primary and backup paths, favouring resource sharingbetween the current demand and the already established requests.After weights are assigned to the network links, Dijkstras algorithmis used to calculate the shortest path to find the primary path. If nosuitable path (based on BER) is found, the connection is blocked.After selecting the primary path, the process is repeated to selectthe backup path, but the primary path is removed from the process.Random pick, first-fit or last-fit schemes are used to select a pathfrom the list of candidate paths.

Experiments and/or analyses conductedExperiments were conducted and results generated, by simulations,on the Pan-European test network defined by COST 239 that com-prises of 11 nodes and 26 bidirectional links each of capacity of 16wavelengths. The behavior of last-fit, random-pick and first-fit forbackup path were investigated. Physical layer constraints were con-sidered separately.

Results that the authors claimed to have achievedThe authors claim that their results show that last-fit wavelength as-signment scheme provides improved network performance of around2 percent when compared to first-fit and significantly outperformsrandom-pick, since it can offer a blocking improvement of 12 percentfor a wide range of traffic conditions. The authors also claim thatnetwork performance is significantly improved when the impairment-aware routing scheme is compared to the impairment-unaware rout-ing scheme, in terms of blocking probability.

Claims made by the authorsThe authors claim that overall network performance can be im-proved when wavelength assignment algorithms, that are able tooffer high spare capacity utilization, are implemented. In addition,they claim that the proposed IAR algorithm incorporated in the


provisioning of protected lightpaths is able to provide guaranteedquality connections in a highly efficient manner.

dos Santos Rosa, S., Drummond, A., and da Fonseca, N.2010. Path protection WDM networks with impaired-transmission.Photonic Network Communications, 19, 2 , 212-222.

The problem which the researcher/authors addressedThe researchers address the problem of protecting impairment-awareWDM networks against failures by focusing on shared path protec-tion schemes that take into consideration various physical impair-ments during path selection.

Previous work by other authors referred to by the authors[Yang et al 2005] [Zhai et al 2007] [Wang and Li 2007]

Shortcomings of previous workThe authors claim that Polarization Mode Dispersion (PMP), Am-plifier Spontaneous Emission Noise (ASE) and cross-talk (CT) im-pairments were not considered in the previous work by other re-searchers.

The new idea, algorithm, architecture, protocol, etc.The researchers developed, what they claim to be, two novel algo-rithms using a shared protection scheme and impairments (PMP,ASE and CT) for the establishment of lightpaths. These algorithmsare First-fit IASPP and Random-selection IASPP. First-fit takes atraffic request and executes a shortest path algorithm to find theprimary path, in order of wavelength. Not considering the primarypath, it then executes the same algorithm to search for a backuppath. If a path is not available the request is blocked.

In the random-selection algorithm, there is no order and the lay-ers are visited randomly. Random selection takes a request andexecutes the shortest path algorithm based on a random selectionof a wavelength. It uses the same methodology for selecting thebackup path, not considering the primary path. If a path is notavailable the request is blocked.


Experiments and/or analyses conductedExperiments were conducted using two network topologies: the NSFwith 16 nodes and 25 bidirectional links and the USA with 24 nodesand 43 bidirectional links. The performance metric used is the block-ing probability.

Results that the authors claimed to have achievedThe authors claim that in the NSF network topology, the resultsof FFIASPP and RSIASSP when compared to FFSSP and RSSSP(without impairment-aware) show: a significant increase in blockingprobability is notices for impaired networks; the FFIASSP reducedthe blocking probability from 0.32 to 0.2, while the RSIASPP re-duced it from 0.3 to 0.17. In terms of vulnerability, the ratios varyvery little.

In the USA network topology, the blocking probability increasedin both aware and unaware algorithms. In terms of vulnerability,the figures were lower than those obtained in the NSF topology.

Claims made by the authorsThe authors claim that results show that this approach decreasesblocking, yet maintains similar level of vulnerability.

Shao, X., Cheng, X., Yeo, Y. K., Zhou, L., Wang, Y.,and Bai, Y. 2010. Performance enhancement for impairment-aware SRLG failure protection in wavelength-routed optical net-works. Photonic Network Communications 20, 1, 1-9.

The problem which the researcher/authors addressedThe researchers address the problem of protecting impairment-awareWDM networks against SRLG failures by selecting working andbackup paths with SRLG-disjoint constraints.

Previous work by other authors referred to by the authors[Zhai et al 2007]

Shortcomings of previous work


The authors claim that in the work of Zhai et al [2007] connectionswith a longer working path or backup path will be rejected.

The new idea, algorithm, architecture, protocol, etc.The new idea considers two policies: Policy I and Policy II. Policy Iis used to satisfy impairment constraints. It computes the two leastimpairment paths. The first path is computed the working path andthen another path is compute for the backup path. Policy II is basedon capacity efficiency. It tries to benefit from both the shortest pathand the least impaired path by choosing them intelligently.

Experiments and/or analyses conductedThe authors conducted two experiments (simulations), one usinga 14-node topology and the other using a 28-node topology. Theycompare the Policy I and Policy II based physical impairment- awareshared-path protection (SPP) with impairment-unaware SPP thatdoes not consider physical impairment.

Results that the authors claimed to have achievedThe authors claim that the results show that the two impairment-aware methods significantly outperform impairment-unaware SSPwith the presence of physical impairment constraints. Also, theyclaim that the results show that generally, the impairment-awareSPP based on Policy II outperforms that based on Policy I in termsof blocking probability.

Claims made by the authorsThe authors claim that when compared with impairment-unawareSRLG failure protection, impairment-aware performs much better interms of blocking probability especially with strong physical impair-ment constraints. They also claim that impairment-aware SRLGfailure protection can significantly reduce physical-layer blockingprobability, and the algorithm based on Policy II achieves a goodbalance between capacity efficiency and physical impairment re-quirements. In addition they claim that the problem is NP-complete.

Kokkinos, P., Manousakis, K., and Varvarigos, E. 2010.Path protection in WDM networks with quality of transmission lim-


itations. Communications (ICC), 2010 IEEE International Confer-ence, 1-5.

The problem which the researcher/authors addressedThe researchers address the problem of protecting impairment-awareWDM optical networks against failures by selecting the primarypath and the backup path, jointly.

Previous work by other authors referred to by the authors[Yang et al 2005] [Zhai et al 2007] [Azodolmolky et al 2009]

Shortcomings of previous workThe authors claim that in the work of Yang et al [2005] only con-sidered two linear constraints (PMD and ASE) and ignored all theother physical constraints. They also claim that in the work of Zhaiet al [2007], keeping the backup path lit worsen the impairmentsfor other lightpaths and keeping the backup path dark can lead toincreased traffic restoration time. They further claim that the workof Azodolmolky et al [2009] was limited to transparent all opticalnetworks and only considered dedicated path protection for offlinedemands.

The new idea, algorithm, architecture, protocol, etc.The authors developed a multicost approach that they claim is sig-nificantly more powerful than single cost routing. A vector of costparameters, which includes the utilization of wavelengths, is assignto each link. These cost parameters are used to calculate the costvector of a path. In this multicost approach more than one path iscalculated between two nodes and, as such, minimization and max-imization can be used to arrive at the best path. This approachconsists of two phases. The first phase uses an algorithm that isa generalization of the Dijkstras algorithm which obtains a set ofcandidate non-dominated lightpaths for a given connection request.In the second phase, a primary lightpath is selected to serve theconnection request while a second lightpath, from the same set isselected as the backup path.

Experiments and/or analyses conductedThe authors conducted simulation experiments considering an all-


optical network where connections arrived dynamically and have tobe served as they come. The blocking probability of several mul-ticost based algorithms as a function of the number of availablewavelengths, performing 1+1 and 1:1 protection were measured.

Results that the authors claimed to have achievedThe authors claim that the results show that in case of 1+1 protec-tion, the sumQ and bestQ-bestQ policies result in similar blockingprobability and in the case of 1:1 protection, the differences betweenthe different protection policies are minimized. They also claim thatthere is a level of trade off between the level of protection (1+1, 1:1and no protection). In addition, the authors claim that the usage ofthe worse case interference assumption for the selection of backuplightpath reduces the blocking probability.

Claims made by the authorsThe authors claim that, based on their experimental observations,the joint selection of the primary and backup lightpaths is generallymore preferable than selecting them separately. They also claim toshow that the selection of backup lightpath, based on the worse caseinterference assumption can be quite beneficial for the survivabilityof the backup paths, when the network is heavily loaded. In addi-tion, they claim that the problem is NP-complete.

Georgakilas, K., Katrinis, K., Tzanakaki, A., and Mad-sen, O. 2010. Performance evaluation of impairment-aware routingunder single- and double-link failures. Optical Communications andNetworking, IEEE/OSA Journal 2, 8, 633-641.

The problem which the researcher/authors addressedThe researchers address the problem of protecting impairment-awareWDM networks against failures by focusing on performance evalua-tion in impairment-aware routing under single and double link fail-ures in identifying the primary and backup paths.

Previous work by other authors referred to by the authors[Markadis and Tzanakaki 2008]


Shortcomings of previous workThe authors claim that the work of Markadis and Tzanakaki [2008]did not take into consideration double-link failures.

The new idea, algorithm, architecture, protocol, etc.The authors developed a modular approach that divides the prob-lem into several modules. The RWA problem is solved in two steps:routing and wavelength assignment (RWA) for the primary path andRWA for the backup path. Both steps are implemented using Di-jkstras algorithm. The primary path for each request is computedby the impairment-aware RWA scheme and uses a first-fit algorithmbased on the quality of the paths (a predefined BER threshold) forwavelength assignment. The backup path is computed using IA andminimum-hop routing and uses the weighted last-fit algorithm forwavelength assignment (also based on the BER threshold)

Experiments and/or analyses conductedFour Experiments, based on simulations were evaluated using twonetwork topologies: the COST 239 topology and the NSFNET topol-ogy. The scenarios are shown in the table below:

Scenario Backup Path Routing Reinforced Sharing

Sc1 IAR Yes

Sc2 IAR No

Sc3 min-hop Yes

Sc4 min-hop NoTable 1: Evaluation Scenarios Definitions, Performance evaluation of impairment-

aware routing under single- and double-link failures [2010]. Georgakilas et al, page

637.

Results that the authors claimed to have achievedThe authors claim that they observed that when IAR was appliedfor the backup path computation, the total blocking probability forthe COST 239 topology was much lower compared with the min-hop; for all load conditions, the blocking probability is reduced bymore than 50 percent. In the NSFNET, the IAR also demonstratedreduction in blocking probability. They also claimed that when re-inforced sharing is applied there is better resource utilization.

Claims made by the authors


The authors claim that when the quality of the formed paths istaking into account in the process of routing, the physical perfor-mance of the selected paths is shown to be better, compared withthe traditional routing schemes that considers only bandwidth avail-ability. They also claim that the impairment-aware routing schemeprovides better performance under single- and dual-links failures. Inaddition, they claim that the consideration of impairments in therouting algorithm for both the primary and backup paths providesa reduced network blocking probability.

Patel, A., Jue, J., Wang, X., Zhang, Q., Palacharla,P., and Naito, T. 2010. Survivable impairment-aware trafficgrooming with dedicated path protection. Optical Communication(ECOC), 2010 36th European Conference and Exhibition, 1-3.

The problem which the researcher/authors addressedThe authors address the problem of protection of impairment-awaretraffic-grooming WDM networks against single failures by focusingon efficient placement of equipment.

Previous work by other authors referred to by the authorsThe authors did not refer to any previous work that attempt to solvethe problem.

The new idea, algorithm, architecture, protocol, etc.The authors developed a ROADM architecture that places a traf-fic grooming equipment or a regenerator at each node. They alsodeveloped a heuristic algorithm that is based on an auxiliary graphthat consists of physical links (physical network topology), auxiliarylinks (routes that satisfied the impairments constraints) and virtuallinks (established lightpaths with sufficient spare capacity for therequest).

A new backup lightpath is established by finding the shortest pathon the auxiliary graph that is a link disjoint from the working path.Regenerators are then placed at all intermediate nodes along thebackup path. Also, if there is spare capacity, the algorithm estab-lishes a separate working lightpath and a separate disjoint backup


lightpath.

Experiments and/or analyses conductedExperiments were conducted, via simulations, on a 100-node meshednetwork topology in which each link is assumed to have a bidirec-tional fiber. Line rates of 10 Gbps, 40 Gbps and 100 Gbps withdifferent optical reach were used in the simulations.

Results that the authors claimed to have achievedThe authors claim that their results show that the cost of the grooming-only approach is less than the cost of the regenerator-only approachand the cost of the combined traffic grooming and regenerator place-ment is always lower. They also claim that based on the results: atlow load, the 10 Gbps network has the lowest cost; at moderate load,the 40 Gbps network has the lowest cost; and at high load, the 100Gbps with reachability of 8 hops has the lowest cost.

Claims made by the authorsThe authors claim that the efficient placement of traffic groomingand regenerators together for establishing survivable lightpaths re-duces cost. They also claim that as the traffic load increases, thenetwork can further be improved by increasing the line rate.

Azodolmolky, S., Klinkowski, M., Pointurier, Y., An-gelou, M., Careglio, D., Sol-Pareta, J., and Tomkos,I. 2010. A novel offline physical layer impairments aware RWAalgorithm with dedicated path protection consideration. Journal ofLightwave Technology, 28, 20, 3029-3040.

The problem which the researcher/authors addressedThe researchers address the problem of protecting impairment-awareWDM networks against failures by focusing on offline physical layerimpairment aware routing with dedicated path protection consider-ation.

Previous work by other authors referred to by the authorsThe authors do not refer to any previous work that attempts tosolve the problem.


The new idea, algorithm, architecture, protocol, etc.In this performance evaluation, the authors developed three algo-rithms. They enhanced two existing IA-RWA algorithms (the RS-RWA algorithm and an ILP-based RWA algorithm) that did notconsidered dedicated path protection. The two offline IA-RWA al-gorithms were extended to consider dedicated path protection. Theperformances of these two algorithms were compared with that of anew IA-RWA algorithm (the Rahyab algorithm). The new algorithmuses a demand pre-processing module to order the traffic demandsand a process to select a lightpath form the candidate lightpath list.Traffic demands can be protected or unprotected.

Experiments and/or analyses conductedExperiments were conducted based on simulations that used theDeutsche Telekom National Network (DTNet) which 14 nodes and23 bidirectional links, with an average node degree of 3.29.

Results that the authors claimed to have achievedThe authors claim that their results show that the performance ofthe ILP-RWA-LU (1 SP) is better than the original ILP-RWA al-gorithms (both 1 SP and 2 SP). They also claim that their resultsindicate that increasing the number of candidate shortest paths alsoincreases the blocking rate of ILP-RWA and ILP-RWA-LU algo-rithms.

Claims made by the authorsThe authors claim that their enhancements to the RS-RWA heuris-tic (that is, RS-RWA-Q) reduced the blocking rate of demands by anaverage of 35 percent from different loads compared to the originalscheme (RS-RWA); the performance of ILP-RWA-LU (1 sp) formu-lation (enhanced IRL-RWA) is also improved by 71 percent; andthe Rehab algorithm performs better than the two algorithms withrespect to the blocking rate performance metric.

Gao, C., Cankaya, H., Patel, A., Jue, J., Wang, X., Zhang,Q., Palacharla, P., and Sekiya, M. 2012. Survivable impairment-aware traffic grooming and regenerator placement with connection-


level protection. Optical Communications and Networking, IEEEOSA Journal. 4, 3, 259-270.

The problem which the researcher/authors addressedThe problem identified by the authors is protection of impairment-aware optical WDM networks against failures by focusing on im-proving utilization of network resources.

Previous work by other authors referred to by the authorsIn 2010, a subset of the authors proposed a combined traffic groom-ing and regeneration placement approach to support all traffic de-mands with minimum cost. [Patel et al 2010].

Shortcomings of previous workThe authors claim that the solution of Patel et al [2010] may notbe cost or resource efficient for cases in which the lightpath is notfully occupied. The new idea, algorithm, architecture, protocol, etc.The authors developed a reconfigurable optical add-drop multiplexer(ROADM) node architecture. They state that this architecture hasan all-optical wavelength switch fabric coupled with an electricalbackbone, which together connects the various grooming and regen-erator equipment.

The authors present two algorithms (a dedicated connection-levelprotection algorithm and a shared connection-level protection algo-rithm) that use auxiliary graphs that take into consideration theweight and residual capacity of each type of link. They note that,in each algorithm, the auxiliary graph is first generated and used toselect the solution that generates the minimum network cost.

Experiments and/or analyses conductedThe authors conducted experiments for both algorithms, on two sim-ulated networks. A randomly generated 100-node network topology,whose average node degree is 4, and a 14-node NFS network with anode degree of 2.7.

Results that the authors claimed to have achievedThe authors claim that their results show that when comparingconnection-level protection with lightpath-level protection, among


other parameters, connection-level protection achieves lower cost,but has a higher signaling cost when a lightpath failure occurs.

Claims made by the authorsThe authors claim that their stimulation results show that their ap-proach outperforms a lightpath-level protection approach, in termsof network cost. They also claim to show that a network with higherline rate and longer optical reach can reduce the network cost, whenthe traffic load increases.

path protection in impairment-aware wdm networks

Documents