List of Abstracts

 

Transparent Optical Networks

 

Raghu Ranganathan

CIENA Corporation

An overview will be given on transparent optical networks.

 

FTTH PON – The Next Big Thing in Telecom

 

André Girard

EXFO Electro-Optical Engineering Inc., Québec

Fiber-to-the-home (FTTH) Passive Optical Networks (PON) is currently in major deployment in Asia, but has also seen steady and strong growth in USA since 2003. There are now demonstration and deployment projects all over the world. The technology is currently being supported by international standards organizations such the International Telecommunication Union (ITU-T) and the Institute of Electrical and Electronic Engineers (IEEE). We will review the market and business motivation and growth figures, and the existing major players involved in such a huge development. We will look at the various available topologies and system architecture optimized for the housing and enterprise development projects. We will at last consider the various system specifications and technical challenges based on practical considerations. A PON Guide will be distributed at the end of the presentation.

 
Topological Design and Dimensioning of Agile All Photonic Networks
 

A.L. Vinokurov, N. Zhao, L. Mason, and D.V. Plant

McGill University, Montréal, Canada

We present the design and analysis of an Agile All-Photonic Network (AAPN); in the context of our study, the agility is derived from sub-microsecond photonic switching and global network synchronization. We have articulated a set of circuit design alternatives in terms of switch configurations referred to as symmetric and asymmetric designs, and 2-layer and 3-layer designs and discuss the implications of these alternatives in terms of transmitter and receiver design and synchronization requirements. In order to evaluate performance and cost of this range of design alternatives, we developed a set of software tools and methodologies for designing and dimensioning our vision of an AAPN. The topological design problem consists of determining the optimal number, size and placement of edge nodes, selector/multiplexers and core switches as well as the placement of the DWDM links so as to minimize network costs while satisfying performance requirements of the supported traffic. A new mixed integer linear programming formulation is presented for core node placement and link connectivity. A methodology has been developed for two-layer and three-layer network topology design and implemented in software. These tools were exercised under a wide variety of equipment cost assumptions for both a metropolitan network  and a long-haul network assuming a gravity model for traffic distribution and a flat community interest factor. Key findings include the determination of near cost optimal designs for both metropolitan (two-layer design) and a Canadian Wide Area Network (WAN, three layer design). We also show the cost and topology sensitivity to the selector switch size and the preferred size in terms of port count and number of switches.

 

 

Recent progess in reconfigurable elements : an optimization perspective

 

Véronique François

Ëcole de Technologie Supérieure, Département de Génie Électrique, Montréal, Canada.

What are the benefits of a reconfigurable WDM optical network? Which are the reconfigurable network elements available today for novel architectural planning and which new optimization do these advanced capabilities open up? These are the questions that the presentation will address. Recent progress of a number of advances optical devices and subsystems will be outlined. Of special interest are the recent developments in the wavelength-tunable and the spectrally-shaped transmitters, which respectively solve an inventory cost issue and the tough problem of dispersion compensation in link budgeting, and may as well enable customer wavelength interchangeability to some extent. Reconfigurable optical add/drop multiplexer and optical cross connect designs using optical Micro-Electronic-Mechanical Systems are becoming available to provide a solution to the challenge of switching/routing the large number of wavelength channels present on WDM backbones. Agile wideband optical amplifiers and tunable dispersion compensators will also be discussed. Moreover, an overview of the ambitious reconfigurable fiber optic test-bed that is being developed at École de technologie supérieure will be presented. We will discuss the new research opportunities and partnerships it will permit for the development and the functional evaluation of intelligent networks.

 

 

Advance light-path provisioning in inter-domain optical networks

 

H. Hakim

Université de Montréal

 

M.G. Khair

School of Information Technology and Engineering (SITE) University of Ottawa

 

A. Maach

Université de Montréal

 

In inter-connected optical networks, the major issues are bandwidth management and fast service provisioning. The goal is to provide optical networks with intelligent networking functions and capabilities in its control plane to enable rapid and guaranteed optical connection provisioning.

 

Users submit connection requests at the time they wish to establish a connection. The service provider consults the information gathered by the inter-domain routing protocols for available resources.  For each request, the network must decide immediately whether to accept or reject the request.  In this model, there is always the uncertainty of whether the user will be able to establish the desired connection at the desired time or not. In fact, the user is provided with no choice which is either acceptance or rejection.

 

The objective of this paper is to develop an advance scheduling technique for inter-domain routing for wavelength-routed optical networks. This advance reservation scheme allows the user to schedule and reserve the necessary resources for a future connection. The user requests a wavelength [startTime, startTime + length] between source and destination that may be located in different Autonomous System s (AS). As a response, the service provider will confirm the reservation with the user before the transmission start time. Therefore, the user is “guaranteed” that the light-path will be established at the desired time. However, if the user requirements cannot be met because of shortage in resources,  the user and the service provider will  have enough time to engage a  renegotiation to reach a mutual agreement on an another starting time.

 

At each border node, the routing table includes the available wavelengths, at different times in the future, for each destination. This information is updated periodically or based on the number of changes that took place in the table. Despite the fact that the routing information is refreshed periodically or when the changes threshold is reached, there is no guarantee that routing information is up to date (e.g., each node waits for its threshold of change in the link table before it advertises). On the other hand, although this routing information could not be very accurate, it still could be considered as a starting point of investigating a possible reservation. Indeed, we use this information to determine the availability of wavelengths to setup a light-path at the requested time (in the future). If we determine that a light-path can be setup, a setup request propagates along the path on its way to the desired destination. If the chosen wavelength is available along the whole path a confirmation message will be send by the destination node back to the source; this confirmation message will be used to reserve the resources, starting from the requested start time, along the path. Otherwise, another wavelength and path will be selected until a light-path is successfully established at the requested start time or at a later time (upon user acceptance).

 
Customer Application Throughput Behaviors facing various Restorations Scenarios on ASTN/ASON
 

Claude Poulin. Hugues Tremblay

Bell

Several IP-based customer applications can withstand some ASTN/ASON recoveries as robustly as traditional Ring based networks with minimum performance degradation.  This presentation explores traffic/restoration behaviors to provide some facts to help design the meshed Optical networks of the future.

 

p-Cycles: Ring Speed with Mesh Efficiency and Extensions into Dynamic Provisioning, Node Recovery and Path Protection


Wayne D. Grover

TRLabs, University of Alberta

 

For most of the last decade there was a divide between the ring and the mesh-based approach to transport network survivability. Rings offered the advantages of speed, structural simplicity, and the determinism of the rerouting process. But the mesh-based approach offered better spare capacity efficiency and the flexibility of handling growth, changes, and multiple classes of survivability.  The authors recently developed technique of p-cycles combine all these advantages in a single scheme for survivability. As such p-cycles are under increasing study as one of the most attractive new technology options for carriers in the next generation of transport networks. 

In this talk we will review the origins of p-cycles in basic research about pre-cross-connection strategies and describe the basic features and operation of p-cycles, and their related design models. We touch on the special role of Hamiltonian cycles as p-cycles, and show that a Hamiltonian p-cycle can by construction reach the lower limit in redundancy for any span restorable mesh technology.  We then survey a number of the most important extensions and new insights about p-cycles including techniques for node as well as span protection, techniques for handling dynamic protected service provisioning, and an important recent extension of the entire p-cycles concept to provide fully pre-connected path protection. The relationship of p-cycles to other recent work on oriented cycle covers, pre-cross-connected trails (PXTs),  and pre-cross-connected trees will also be explained.

 

 

Combinatorial Approximation Algorithm for the Integral Multicommodity Flow Problem

 

Hervé Rivano

INRIA Sophia-Antipolis, France

 

The design of WDM optical networks is an issue for telecommunication operators since the spreading of this technology will not occur unless enough performance guarantees are provided. Motivated by the seek for efficient algorithms for the Routing and Wavelength assignment problem (RWA), we address approximations of the fractional multicommodity flow problem which is the central part of a complex randomized rounding algorithm for the integral problem. Through the use of dynamic shortest paths computations and other combinatorial approach, we improve on the best known algorithms. We also provide directions for further improvements.

 

Branch and Price for WDM Optical Network Design

 

S. Raghavan and Daliborka Stanojević

University of Maryland, USA

Optical networks with wavelength division multiplexing represent the second generation of optical networks designed to take advantage of large bandwidths available in these networks. Efficient utilization of these networks, however, requires the simultaneous design of the logical topology, where we need to construct lightpaths over the physical topology of the optical network, and routing of traffic over the established logical topology. In a sense, this problem can be seen as a two-layer integer multicommodity flow problem, a more complex version of the standard integer multicommodity problem that has been efficiently solved using branch and price and cut algorithms in the literature. The more complex side of the two-layer integer multicommodity flow problem lies in the fact that both the number of lightpaths, and the traffic paths defined on these lightpaths grow exponentially with the network size. Consequently, we propose a column generation based algorithm that starts with only a subset of lightpaths and traffic paths and then progressively adds new paths to the model. We present our preliminary computational results and discuss the applicability of our algorithm to the design of optical networks for several alternative design objectives and requirements.

 

MDP Based Link Cost Optimization for Protection in Optical Networks

 

Zbigniew Dziong and Hadda Khedimi

École de Technologie Supérieure, Montréal, Canada


Ramesh Nagarajan

Lucent Technologies – Bell Labs, Holmdel, NJ, USA

 

In today’s competitive market, network resiliency to failures has become an important feature that is used to attract new customers. At the same time this feature has to be realized in a cost effective fashion due to current industry financial constraints. In the paper we address the issue of optimal resource management in meshed optical networks with shared path or link protection. We formulate the problem of primary and protection path calculation as a Markov decision problem. After decomposition of the problem, we define the dynamic link costs as state dependent link shadow prices that correspond to predicted lost revenue due to the new connection admission. While exact solution for link shadow prices can be quite complex, we use the Markov decision formulation to derive an approximate relation between the link costs for primary and shared protection connections. This result is then used to formulate an efficient link cost structure that is applied to optimize the primary and protection paths from the bandwidth usage viewpoint. Numerical results show the relation between link costs for primary and protection connections. Also the gain achieved from link cost optimization is illustrated.

 

Survivable Traffic Grooming in Optical Networks with Multiple Failures

 

Wei Huo and Chadi Assi

Concordia Institute for Information Systems Engineering

Concordia University, Montreal, Canada.

 

 

Routing connections with diverse bandwidths and protection requirements in large capacity optical networks is defined as the survivable traffic-grooming problem. Recent studies have focused on defining different grooming policies and have proposed different protection frameworks against single element failure. However, as the size and the complexity of multi-granular optical networks continue to grow, dual-failures become increasingly probable. Various research efforts have addressed the problem of routing high-capacity lightpaths1 under dual-failures assumptions, and findings show that designs offering complete dual-failures restorability require more than double the amount of spare capacity. Moreover, none of the proposed approaches are applicable to multi-granular networks due to their associated complexity. Capacity re-provisioning, alternatively, exploits the available capacity in the network to re-establish new protection paths for unprotected connections right after the recovery from the first fault. In this paper, we propose to protect multi-granular optical networks against near-simultaneous dual-failures using capacity re-provisioning and we study the performance of reprovisioning under different protection frameworks; namely, lightpath level and connection level protection. We study the performance of reprovisioning through simulation experiments, and we show that the restorability of groomed optical network against dual failures can substantially be improved.

 

Wavelength Tracking Tools for User Empowered Reconfigurable Optical Networks

 

Mathieu Lemay, Michel Savoie

Communications Research Center, Ottawa, Ontario, Canada

 

Véronique François

Département de Génie Électrique, École de Technologie Supérieure, Montréal, Canada

 

DWDM reconfigurable network elements now provide dynamic optical networks but a major problem associated with these networks is a need to keep track of the different wavelengths used and to know what path specific data is taking. This presentation proposes a midterm solution for wavelength tracking and performance monitoring for today's reconfigurable optical networks. It demonstrates how a sensor based network consisting of software services is capable of providing tools for user empowered all optical networks using out of band signalling. This system can also be used by different User Controlled Lightpath Provisioning (UCLP) implementations to monitor changes on lightpaths or to provision routes in all optical networks (AON). Furthermore, network element intelligence can be added to provide plug and play behaviour as well as static power (OSNR) optimization. 

In this talk, we present the discovery process of different elements along physical fibers as well as the signalling process needed to create a lightpath surrogate. Also, the capabilities of this surrogate and its potential use for RWA or by UCLP will be explained. Finally, a brief description of all the information gathered by monitoring functionalities will be presented.

 

 

Network cost Impact of Solutions for Mitigating Optical Impairments:
Comparison of Methods, Techniques, and Practical Deployment Constraints

Michel P. Bélanger

Nortel Networks

 

The network costs of dispersion compensation strategies are reviewed. Practical field issues such as PMD, non-uniform span loss distributions and OADM placement are considered.  The performance and cost impact of electrical and optical methods are compared.

 

 

Tools for the Design of All-Optical Networks


Hussein T. Mouftah

School of information Technology and Engineering
University of Ottawa, Ottawa, Ontario

 

The ever increasing demand for transmission bandwidth, driven by the growth of the Internet and related services, is stretching the capabilities of existing technology for IP routers, especially in the high capacity Internet backbones. Incoming wavelength division multiplexing (WDM) technology provides a platform to exploit the potential huge capacity of the optical fibers. WDM optical networks are poised to dominate the backbone infrastructure supporting the next-generation high-speed Internet backbones. The challenge now is to combine the advantages of WDM with emerging all-optical switching capabilities to yield high throughput optical routers. In this presentation, we will address these issues and discuss ways that are being sought to maximize the benefits of optical networks while minimizing costs. We will also discuss the role wavelength conversion may play to improve efficiency in all-optical networks

 

Sharing the bandwidth of a wavelength: Burst switching with and without reservation

 

Gregor Bochmann

University of Ottawa

We propose optical burst switching with various retransmission schemes to recover burst losses due to contention. We consider a star topology network. Our simulation studies compare these different retransmission schemes and also provide a comparison with burst switching with advanced reservation which avoids contention losses completely. We will present comparative values for different performance parameters, such as average effective transmission delay and bandwidth utilization, for varying network parameters, such as number of edge nodes, traffic load, and propagation delays between the edge and core nodes.

 

Packet aggregation at the optical network edges


Gérard Hébuterne

INT, Evry, France

In optical packet networks, technological constraints at the physical layer favour the choice of constant duration packets, the order of magnitude being microseconds – i.e. several thousand bytes, assuming e.g. 40 Gbit/s wavelengths. It amounts to impose an aggregation of user packets, as IP packets have much lower lengths (40 % are around 60 bytes long). 

The aggregation device works as an interface between the electronic and optical domains: it receives IP packets (or ATM cells, etc.) from the client layer and converts them into optical packets to be sent inside the optical network. To reduce the aggregation delay, a timer is added to the aggregation mechanism, in such way that an optical packet is transmitted whether the timer expires or the aggregation unit becomes full or cannot accommodate more IP packets.

The purpose of the talk is to present an analytical model, based on discrete time Markov Chains, which incorporates the timeout mechanism.  We develop an exact mathematical model for the performance evaluation of packet aggregation on the edge router of an optical network. We derive closed formulas for the probability density function of the aggregation delay, the optical inter transmission time, and of the transmitted optical packet length. Performance measures such as aggregation delay and efficiency show the importance of the timer in bounding the delay of transmitted packets without much altering the aggregation efficiency.

 

Time-slot reservation in all-photonic networks based on flow prediction

 

Mark Coates

Dept. Electrical and Computer Engineering

McGill University

In wide-area all-photonic networks that rely on confirmed reservation for control of core optical switches, it is necessary to make the reservations a significant amount of time before the arrival of the traffic in order to allow for round-trip delays. This necessitates prediction of future requirements of source-destination traffic flows. In this presentation, we will focus on all-photonic networks that have an overlaid-star topology. We will describe an approach based on traffic sampling and distributed expectation-maximization for predicting the resource requirements of end-to-end flows traversing the network. The estimates provide the foundation for a bandwidth allocation scheme based on fixed-length time-slot reservation. We study the performance of this allocation technique and compare it with other approaches.

 

User Controlled LightPath : UQAM Approach

 

Omar Cherkaoui, Halima Elbiaze, Boubker Ghandour

Department of Computer Science, UQAM, Montréal, Canada

The UQAM UCLP (User Controlled LightPath) software is designed to allow end users of multi autonomous systems (domains) to create lightpaths for their own specific applications, particularly to satisfy the needs of high end greedy applications. For example a community of high energy physicists’ researchers can create their own lightpaths over a multi-domain network in order to share their local resources and make them accessible for the other community members based in different domains (laboratories, universities, cities, countries …).

UQAM UCLP enables users to reconfigure a large optical network and build an “independent IP network”, then define its topology and architecture that are optimized for the needs and requirements of their particular applications. The owners of these networks are allowed to reconfigure their independent IP networks at any time without getting the permission of the domains managers.

In the presentation, we will present the following: the constraints resolved by the UQAM UCLP implementation, the security (users Authentication and Authorization), the synchronization with network resources (inventory), the inter/intra domain constraints (computing routes). WE will also present the UQAM UCLP architecture with an illustration of each module, as well as the UQAM UCLP signaling process to provide the lightpath provisioning service.

 

Demonstration will be possible during one of the coffee break for the interested participants.

 

Dynamic Bandwidth Allocation in TDM/WDM-PON

 

Ahmad R. Dhaini, Chadi M. Assi

Concordia Institute of Information Systems Engineering,

 

Abdallah Shami

ECE Department, University of Western Ontario

 

IEthernet Passive Optical Networks (EPON) have emerged as one of the most promising access network technologies. Currently, various Dynamic Bandwidth Allocation (DBA) algorithms have been implemented to arbitrate the transmission of multiple users over the shared single channel and to improve the overall performance (packet delays, jitter, packet loss, etc.). Current EPONs are cost effective but their bandwidth is limited; hence they do not scale to support a large number of Optical Network Units (ONUs) as well as the continuous increase in the demand for bandwidth. Now, given the tremendous success of Wavelength Division Multiplexing (WDM) technology in backbone network, it is natural to deploy this technology in the access network to achieve high performance. In this presentation, we discuss the integration of TDM-PON with WDM and we expand the Medium Access Control (MAC) to support both time and wavelength allocation. We extend the requirements of the Optical Line Terminal (OLT) DBA to support the management of both time and wavelength. Finally we present some simulation results on TDM/WDMPON DBA performance, and our future proposed work.

 

Optimized Architectures for IP / Optical Carrier Grade Networks

 

Thanh Huynh

Bell Canada – Network Planning

 

IP is becoming the dominant force as the communication infrastructure in the 21st century. And it is now well accepted that IP will be the foundation for communications worldwide for almost all applications for both business and residential. This paper presents various IP / Optical architectures that allow IP to fill this important role by providing the carrier grade reliability at affordable cost.
Some well documented approaches based on mesh networks require a paradigm shift from current mode of operation. Other design scenarios, still unexplored based on incremental enhancements of existing architecture, use recent developments in IP / MPLS - TE and QoS to unlock the latent capacity of SONET rings.

 

 

 

 

Traffic Grooming (Multiplexing) and Inverse Multiplexing Problems in Telecom Networks

 

Biswanath Mukherjee

Department of Computer Science, University of California

 
 

Traffic grooming is the problem of efficiently multiplexing (packing), demultiplexing (unpacking), and switching low-speed tributaries onto high-capacity channels, e.g., wavelengths or lightpaths in optical wavelength-division-multiplexing (WDM) networks.  While SONET/SDH (over WDM) is the dominant telecom infrastructure, Ethernet is the ubiquitous data-network interface; thus, next-generation SONET/SDH technologies have been developed to efficiently carry data over our telecom infrastructure.  In this talk, we will discuss how Data/Ethernet over SONET/SDH over WDM works and how it leads to new and exciting research problems involving inverse multiplexing, multi-path routing, provisioning/reliability/survivability of service paths, and network design.

 

Survivable Optical Network Design by Mathematical Optimization

 

Arie Koster

Konrad Zuse Zentrum für Informationstechnik, Berlin, Scientific Computing - Optimization

 

The potential of mathematical optimization techniques for the cost-efficient design of survivable optical networks is the topic of this talk. We show how the optical network design problem can be modeled as an integer linear program, how the consideration of wavelength converters allows for a decomposition approach to find good network configurations, how the quality of these network configurations can be benchmarked, and how survivability in transparent optical networks can be implemented by the optimization-oriented concept of Demand-wise Shared Protection.

The Input of optical network design consists of (i) a network topology, (ii) a traffic network which specifies the demand for each pair of nodes as number of lightpaths to establish, (iii) a survivability matrix which specifies for each pair of nodes the number of connections that should survive any node or link failure, and (iv) the existing and purchasable (optical) devices: fibers, WDM-systems, OXCs, regenerators, and wavelength converters. The objective is to design a minimum cost (multi-hop) transparent optical network configuration such that the installed devices provide enough capacity to establish a set of lightpaths in a conflict-free manner, which on their turn suffice to handle the traffic in all the cases in which at most one node or link fails.

A mathematical problem analysis exhibits that the problem is too complex to handle by state-of-the-art integer programming techniques, since it contains both the integer multi-commodity flow and the vertex coloring problem as sub-problems. Therefore, we use the availability of wavelength converters to decompose the problem into a dimensioning and routing part and a wavelength assignment part.

For the dimensioning and routing sub-problem, we present an integer programming based solution approach, including pre- and post-processing tasks to reduce the complexity. Besides a dimensioning of the network and routing of the lightpaths, the approach also provides a lower bound on the total network cost that is inevitable, given the prerequisites. By this bound, the quality of the solution of the network configuration can be evaluated.

For the wavelength assignment subproblem, we briefly discuss some heuristic algorithms. Moreover, we derive a combinatorial lower bound on the number of converters by generalizing the theory of edge coloring.

To guarantee that also in case of a node or link failure the sustainable traffic is accommodated, we discuss a recently introduced concept for the ressource-efficient design of survivable meshed optical networks. Since it applies protection sharing within single demands, it is called Demand-Wise Shared Protection (DSP). It is in particular suited for meshed optical networks, taking advantage of the connectivity in meshed topologies as well as taking care of the specific routing requirements imposed in all-optical networks.

We show that in comparison with the well-known 1+1 protection concept, the number of required backup lightpaths may reduce significantly. Moreover, DSP can be easily incorporated into the integer programming formulation for dimensioning and routing, whereas the wavelength assignment is not affected. From the reduced number of required backup lightpaths, it follows that substantial network design cost reductions can be achieved. Computational experiments approve these findings.

Most of the presented results are joint work with Adrian Zymolka (ZIB). This work has partially been supported by Telekom Austria and T-Systems Nova.

 

 

A Column Generation Approach for Wavelength Assignment with Converters

 

Arie Koster and Adrian Zymolka

Konrad-Zuse-Zentrum für Informationstechnik Berlin (ZIB)

 

Finding conflict-free wavelength assignments for lightpaths states an important subtask in the cost-efficient design of optical networks. We have given a fixed routing of the lightpaths together with a sufficiently dimensioned hardware configuration comprising fibers, varying WDM systems, and optical switches. On each link, the installed equipment determines the number of times each wavelength is available. To complete the lightpath configuration, we have to assign an available wavelength of operation to each link of each lightpath. We consider the general case of multi-fiber links without allocating the lightpaths to dedicated fibers which would restrict the possible assignments. At the nodes, a lightpath can exchange the operated wavelength at the cost of a wavelength converter. Hence, minimizing the overall network design cost implies minimizing the total number of converters needed.

In this talk, we present solution methods for the Minimum Converter Wavelength Assignment Problem (MCWAP). Sequential and iterative heuristics generate feasible assignments, but lack of a quality guarantee. To benchmark the solutions, we derive exact approaches by use of integer linear programming. We present two different formulations of MCWAP and compare the quality of the lower bounds provided by the linear relaxations to each other and to the integer optimum. The most promising one is a generalization of a formulation by Mehrotra and Trick for vertex coloring in graphs. We develop a column generation method to solve the linear relaxation yielding good lower bounds for MCWAP. In practical cases, optimality of the best known solution could be proven this way. Moreover, we sketch a branch-and-price approach for solving the integer program.

 

 

Traffic Grooming in Simple Networks using Design Theory

 

Jean-Claude Bermond and David Coudert

INRIA Sophia-Antipolis, France

 

In a WDM network, routing a request consists in assigning it a route in the physical network and a wavelength. If each request uses at most 1/C of the bandwidth of the wavelength, we will say that the grooming factor is C. That means that on a given edge of the network, we can groom (group) at most C requests on the same wavelength. With this constraint, the objective can be either to minimize the number of wawavelengths (related to the transmission cost) or minimize the number of Add Drop Multiplexer (ADM) used in the network (related to the cost of the nodes). Here we consider where the network is a unidirectional ring on N nodes, CN, or a path on N nodes, PN. Thus the routing is unique.

For a given grooming factor C minimizing the number of wavelengths is an easy problem, well known and related to the load problem. But minimizing the number of ADMs is NP-complete for a general set of requests and few results are known. Here we show how to model the problem as a graph partition problem and using tools of design theory we solve the case of static uniform all-to-all traffic (requests between all pairs of vertices) for some grooming factors.

 

 

On the Joint Link Dimensioning and Routing Metric Assignment Problem for Reliable WDM Networks

 

Steven Chamberland

École Polytechnique de Montréal, Canada

 

In this paper, we tackle the joint link dimensioning and routing metric assignment problem for reliable wavelength division multiplexing (WDM) networks. This design problem consists in finding the number of wavelength channels on each link and the routing metrics (considering shortest-path routing) that ensure the routing of all virtual wavelength paths (VWPs) and the successful rerouting of the reliable VWPs for all failure scenarios of interest to the network planner. A mixed integer mathematical programming model is proposed for the problem. The model is adapted for the single link failure scenarios. Since the problem is NP-hard, we propose a tabu search algorithm to obtain good approximate solutions for real-size instances of the problem. Finally, a lower bound is proposed and numerical results are presented and analyzed.

 

 

On Column Generation Formulations for the RWA Problem

 

Brigitte Jaumard, Christophe Meyer and Babacar Thiongane

Université de Montréal, Department of Computer Science and Operations Research

 

We present a review of column generation formulations for the Routing and Wavelength Assignment (\RWA) problem with the objective of minimizing the blocking rate. Several improvements are proposed together with a comparison of the different formulations with respect to the quality of their continuous relaxation bounds and their computing solution ease.

 

 

How Important is Scheduling in a Single-Hop Optical Network
 

Akbar Ghaffar Pour Rahbar  and Oliver Yang

School of Information Technology and Engineering (SITE) University of Ottawa

 

Two common classes of methods can be applied to the traffic scheduling in a single-hop all-optical network. The Centralized Schemes schedules collision free slot assignments for the edge switches while the Distributed Scheme gives the edge switches more freedom as when to send the packets.

We have applied these two schemes to the AAPN (Agile All Photonic Network) architecture where propagation delay can be non-negligible. Some performance results are presented and the tradeoffs between these two schemes are discussed in terms of various performance measures such as bandwidth allocation, delay and loss probabilities.

 

Routing in Multigranularity Optical Networks

 

Jean-Marc Hyppolite, Philippe Galinier and Samuel Pierre

Département de génie informatique

École Polytechnique

 

A hierarchy in optical networks with the WDM technology has recently been brought to the scientific community, with promises of cost reduction and a better adaptivity to the network evolutions. A key element in the design of these networks lies in the routing that must groom several optical paths in an efficient way in order to get the best possible avantage in the hierarchy. In this presentation, we will address the routing problem with static requests without wavelength continuity assumptions. We will present a routing algorithm using a Tabou Search heurisitc and we will discuss the numerical results.

 

 

An Optimization problem for WDM Optical Transport Network using Diverse
Routing and Spare Capacity Assignment

 

Ramasamy Mariappan

Sri Venkateswara College of Engineering

Chennai,India.

 

Wavelength division multiplexing optical transport networks are expected to provide the capacity required to satisfy the growing volume of telecommunications traffic in a cost-effective way. New optimization problems arise in connection with these networks. In this paper, we investigate optimization problems arising in the engineering of an optical transport network. Network engineering concerns the configuration of existing network resources in order to satisfy expected traffic demands. A dynamic deterministic traffic model called Scheduled Lightpath Demands (SLDs) is proposed. The model captures the time and space distribution of a set of connection demands and, being deterministic, eases the use of combinatorial optimization techniques to solve network optimization problems. An extensive use of meta-heuristic algorithms is made since they provide approximate solutions of good quality in reasonable computing time to large instances of the investigated optimization problems, which are otherwise computationally intractable. Furthermore, meta-heuristic algorithms ease the introduction of complex constraints found in real-world optimization problems.

 

 

A Performance Comparison Study of OBS and OCS Networks

 

Thomas Coutelen

Department of Computer Science and Operations Research, Université de Montréal, Montréal, Canada

 

Halima Elbiaze

Department of Computer Science, UQAM, Montréal, Canada

 

Brigitte Jaumard

Department of Computer Science and Operations Research, Université de Montréal

 

Optical Burst Switching (OBS) is a new optical switching paradigm where traffic can be switched and groomed at a lower level compared to Optical Switching Circuit (OCS). While OCS is useful in carrying highly aggregated long-lived streams that require absolute Quality of Service (QoS) guarantees, OBS has a role in efficiently carrying bursty best-effort traffic. In this paper, we investigate further the comparative bandwidth efficiency of OBS and OCS using, on one hand, an adaptive deflection OBS routing that aims at minimizing the burst loss rate and, on the other hand, a routing and wavelength OCS strategy using the best possible alternate routing paths for minimizing the blocking rate.