List of Abstracts

Towards the practical deployment of optical code-division multiple-access

Lawrence R. Chen

Photonic Systems Group, McGill University

High-performance computing and communications access networks have several common features: (1) they span lengths of a few tens of meters to hundreds of kilometers, (2) they do not need to be connected to an ultra-high bandwidth long-haul backbone network, (3) they are broadcast-and-select networks, (4) a minimum bandwidth must be guaranteed for a user (e.g. processor unit, intelligent sensor), (5) a given connection may be required to support different data rates. i.e. multi-rate transmission, and (6) asynchronous operation is desired. More importantly, the nature of data transfer will vary from constant to highly peaked (bursty). In order to accommodate these features and meet interconnection specifications, these networks require an access protocol and implementation that can provide the necessary capacity in an efficient manner.

Optical code-division multiple access is a “tell-and-go” access strategy in which data from multiple users are transmitted concurrently over the network and differentiated with code (or user) specific detection. OCDMA does not require the overhead associated with synchronization or network protocols and is well-suited for network applications involving bursty, asynchronous traffic and where aggregate network capacity varies. 

 Although the principles of OCDMA have been known for many years, OCDMA has never delivered on its true potential due to limited system performance arising from fundamental and technological constraints.  In this presentation, we describe some of the key challenges and impediments that have prevented OCDMA from delivering on its potential, discuss possible solutions, and present results of system and technology demonstrators which pave the way for their practical deployment.

Minimum Color Problems and Shared Risk Resource Group in Multilayer Networks

D. Coudert, P. Datta, H. Rivano and M-E. Voge

INRIA Sophia-Antipolis, France

Failure resilience is a desired feature of the Internet. Most traditional restoration architectures assume single-failure assumption, which is not adequate in present day multilayer networks.

 Multiple link failure models, in the form of Shared Risk Link Groups (SRLG's), Shared Risk Node Groups (SRNG's), and more generally shared risk resource groups (SRRG), are becoming critical in survivable optical network design. We develop techniques for the minimum color path problem to tolerate multiple failures arising from a shared resource group failure. We also study the minimum color st-cut problem that we prove NP-complete and hard to approximate. We also provide efficient MILP formulation and heuristic algorithms for it.

QoS Management for Restorable Optical Burst Switched Networks

Salim Y. Said and Hussein T. Mouftah

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

Halima Elbiaze

Optical burst switching (OBS) is a promising technology for next generation optical internet.  A crucial issue for optical burst switched networks is the ability to support a wide range of services, those being supported by IP networks today, and to cope with network failures.  Link failures in OBS networks cause higher burst losses and thus reduce significantly the Quality of Service (QoS) in the network.  We present a QoS-based restoration mechanism for OBS networks based on offset time extensions.   Our simulation study had showed the effect of the offset time value on the network performance and refining QoS levels.  Adding different extension values to the offset time enables the creation of different classes of service correspondent to the data bursts.  This service differentiation guarantees a certain level of data loss and delays in OBS networks under failure.  With the compensation between the level of burst loss and the delay encountered during a restoration process or either parameter alone, one can define a QoS dimension fitting the desired needs.  Simulation studies verified the role of the offset time in the restoration mechanism and its aptitudinal character to define QoS levels.

FTTH Technologies and Testing & Monitoring Issues

Nicholas Gagnon

EXFO, Electro-Optical Engineering Inc., Québec, Canada

Towards the practical deployment of optical code-division multiple-access

Peng He and Gregor Bochmann

Photonic Systems Group, McGill University

In this paper, we study how MPLS flows are routed in an Internet that contains a centrally-controlled agile all-photonic star WDM network (AAPN). Two scenarios are considered, namely deploying AAPN within one OSPF (Open Shortest Path First) area and within several OSPF areas. Since the AAPN provides an NxN interconnection structure for the N edge nodes of the AAPN architecture, the straightforward usage of a routing protocol like OSPF leads to scalability problems. In the first scenario, we have identified several schemes by which this scalability problem can be reduced. The idea is to introduce "virtual routers" that represent a collection of edge nodes (and possibly also the core node), thus reducing the number of paths between the "routers". In the second scenario, we focus on inter-area routing in large-scale IP/MPLS networks. This paper proposes a novel framework for inter-area MPLS Traffic Engineering. The key to our proposal lies in deploying the AAPN architecture as the OSPF backbone area and introducing the concept of "virtual area border routers" (v-ABRs). Compared with other proposals, our proposal can provide globally-optimized inter-area routing and has very good compatibility to existing traditional IP/MPLS routers.

A Simple and Efficient MAC Protocol for All-optical Packet Ring

G. Hébuterne

Institut National des Télécommunications, Evry, France

In all-optical packet networks, rings are commonly envisaged, for collecting traffic from local area networks. The MAC procedure designed for that purpose must be simple and efficient -- and especially its performance must not depend on the traffic pattern on the ring.

The CSMA/CA protocol presents a major drawback, since it leads to an inefficient use of the available bandwidth, raising strong unfair behaviors and lowering the maximum load admissible. This behavior is easily explained, as the protocol favors building a lot of small gaps in the available bandwidth.

We describe a simple modification of this procedure, in which packets are inserted only at the end of already formed "trailers", and we show the improvement, in terms of fairness and bandwidth use.

Recent development of WDM Multiplexing led to a significant increase of transmission capacity broadband networks. While there is now no limitation in the transmission capacity, attention should be now given to the interconnection capacity, i.e., routing and switching capabilities. For this reason, a lot of attention has been devoted to the GRWA - Grooming, Routing and Wavelength
Assignment problem with the objective of either minimizing the capacity or the interconnecting equipment cost. In this study, we include the compensation concerns when solving the GRWA problem and propose a multi-phase Tabu Search to solve the resulting design problem. Numerical experiments were conducted on various network topologies and traffic instances. Results showed that the minimum overall cost is not necessarily reached with the minimum MSPP cost.

Wavelength bandwidth in actual networks ranges up to 10 Gb/s, but bandwidth requests are only a fraction of a wavelength bandwidth. When only one request was assigned to a wavelength in the first generation optical networks, the bandwidth capacity was under-used. In order to optimize the bandwidth utilization, traffic grooming is implemented on today's networks. It consists in multiplexing several requests on a wavelength. However, WDM and grooming techniques require expensive opto-electric systems at nodes, such as Multi-Service Provisioning Platform (MSPP), to perform the operations of grooming, routing and wavelength assignment (GRWA).

We present a review of column generation formulations for the Routing and Wavelength Assignment (RWA) and the Grooming, Routing and Wavelength Assignment (GRWA) problems with the objective of minimizing the blocking rate (RWA) and minimizing the cost (GRWA).  While it is possible to solve realistic network and traffic instances for RWA, it remains a challenging issue for GRWA.

Segmentation-based Random Dropping Scheme in OBS networks

Hongbo Lui and Hussein T. Mouftah

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

As an efficient switching paradigm in WDM layer, OBS is attracting significant attentions from researchers. Contention resolution is one of the most significant concerns in OBS networks. Another concern in OBS networks is QoS differentiation support.

As an important contention resolution approach, segmentation can pass packets using as much as possible fragmented resources. Currently, there are two representative segmentation-based contention resolution approaches: head-dropping and tail-dropping. When contention occurs, they drop either the head of the contending burst or tail of the scheduled burst accordingly. Thus they cannot use bandwidth resources flexibly and efficiently.

Aiming at using the closed intervals between those already scheduled bursts as much as possible, we propose a new segmentation-base contention resolution scheme -- random dropping. When contention occurs, any part of a contending burst may be dropped, instead of only the head or tail of the burst. This dropping policy makes bandwidth utilization more efficient and flexible. Simulation results show that random dropping performs better than head dropping and tail dropping in terms of packet loss rate.

Moreover, in the random dropping approach, we proposed a preemptive and a non-preemptive method. With the non-preemptive method, when contention occurs, it is only the segments from the contending burst that are dropped, and the scheduled burst can be guaranteed not to be dropped before it is delivered to the next node. However, with the preemptive method, the dropped burst is decided based on the product of the size of overlapped segment and its hop number. The preemptive method makes a lower packet loss than the non-preemptive method.

Furthermore, to support QoS differentiation, a new assembly mechanism called composite segment Burstification (CSB) is proposed, which assembles packets of different classes into the same segments. Based on CSB, by adjusting the number of packets of each priority class in segments, QoS differentiation can be achieved.

QoS Management for Restorable OBS Networks

Hussein Mouftah

Canada Research Chair in Optical Networks, School of Information Technology and Engineering, University of Ottawa

Optical burst switching (OBS) is a promising technology for next generation optical internet.  A crucial issue for optical burst switched networks is the ability to support a wide range of services, those being supported by IP networks today, and to cope with network failures.  Link failures in OBS networks cause higher burst losses and thus reduce significantly the Quality of Service (QoS) in the network.  We present a QoS-based restoration mechanism for OBS networks based on offset time extensions.   Our simulation study had showed the effect of the offset time value on the network performance and refining QoS levels.  Adding different extension values to the offset time enables the creation of different classes of service correspondent to the data bursts.  This service differentiation guarantees a certain level of data loss and delays in OBS networks under failure.  With the compensation between the level of burst loss and the delay encountered during a restoration process or either parameter alone, one can define a QoS dimension fitting the desired needs.  Simulation studies verified the role of the offset time in the restoration mechanism and its aptitudinal character to define QoS levels.

Burst Mode Receivers for Agile All-Photonic Networks

Wei Tang, Julien Faucher, Alan Li and David Plant

Department of Electrical and Computer Engineering, McGill University

Recent advances in fiber optic technology have prompted researchers to envision a future all-photonic network that is capable of supporting multiple access and services at very high bit rates. The confluence of optical transmission and optical network functions opens up new paradigms for network architectures that are enabled by emerging photonic technologies. Characteristics of these architectures and technologies that distinguish them from existing ones include: (1) networks in which the transmission of information is based on optical packets (burst-switched or packet-switched networks, with and without all-optical header recognition), (2) optical code-division multiplexing for allocating bandwidth-on-demand in bursty, asynchronous traffic environments, and (3) practical implementations for optical generation, shaping, and processing. In these all-photonic multi-access networks, any node can use a designated time slot to send a packet to any other node in the network. The bursty nature of these networks imposes new design constraints on transmitters, receivers, and optical components. We review various system and technology considerations for receivers in such networks.

Network Connectivity Loss, Impacts on Customer Applications

Claude Poulin and Julie Roberge

Bell

Today's networks have to support a wide variety of IP based services. Customers' needs are becoming more specific in terms of quality of service and customer application performances are today's focus. In that area of research, many traffic engineering techniques are being developed. These latter tend to demonstrate that numerous advantages can be taken from studying more closely network interlayer interactions. More specifically, this presentation highlights key networking behaviours to assess the impact of various failure scenarios on customer applications and demonstrates the efficiency of a well engineered multi layered network (IP / FRR (Fast ReRoute) / MPLS (Multi Protocol Label Switching)  / SONET / optical based Network).

The Role of Engineering and Optimization Techniques in Creating New Services and Service-Level Agreements (SLAs) in Telecom Networks

Biswanath Mukherjee

University of California, Davis
 

Telecom networks -- as well as the services and SLAs they offer -- have typically been built using the "one-size-fits-all" mentality.  However, the underlying network infrastructure is very heterogeneous w.r.t. fiber types, component types, their reliabilities, etc.  In addition, (future) applications generated by the telecom network operator's customers, viz. ISPs, enterprises, and large institutional users of bandwidth, are expected to be quite diverse in their needs, e.g., different service paths require different bandwidth, different protection-switching time, different reliability guarantees, etc.  This presentation will explore the role of engineering and optimization techniques so that a network architect can appropriately match the application needs with the properties of the network infrastructure.  The role of pricing and network economics will also be discussed.

A Simulated Annealing Heuristic for Advanced Lightpath Reservation in WDM Networks

Abdallah Shami and T. Dan Wallace

University of Western Ontario, London Canada
 

Advanced reservation is a topic that is rarely discussed within the domain of wavelength division multiplexed (WDM) networks.  However, for many emerging applications in the telecommunication and/or grid computing industries, a demand for a high bandwidth communication channel as well as a guarantee on resource availability certainly exists.  Such applications include: remote surgery, remote experimentation with teleobservation capabilities, teleconferencing, and bulk transfers.  Furthermore, since internet service providers (ISP) can predict traffic patterns from statistical archives, they will know in advance when and for how long they may require additional bandwidth on selected links.  A solution for these types of network applications can be modeled as a special case of the static lightpath establishment (SLE) problem in which requests not only provide a source and destination, but also a start and end time (i.e. setup and teardown date).  We refer to this new problem as the advanced reservation static lightpath establishment (ARSLE) problem.  In this talk, we first formulate these two new problems as integer linear programs (ILPs) by dividing each wavelength on every link into fixed sized timeslots.  To address the issue of exponential complexity, we have developed a heuristic based on the simulated annealing (SA) algorithm for the problem of maximizing the total number of established lightpaths in advance.  We compared our heuristic to an existing routing and wavelength assignment (RWA) algorithm as well as to the optimal solution.  In terms of the existing RWA algorithm, our SA heuristic yields a significant improvement on the approximate solution by making exponential gains as we increase the number of allowed iterations.  With respect to the optimal solution, our experimental results show that the loss endured by using SA over the ILP will decrease rapidly with smaller values of α, albeit resulting in a poorer solution when compared to a larger value of α that converges more slowly but generates a better approximation to the optimal solution.

Optimization of All-photonic Star Networks

Richard Vickers

McGill University

Computer and Information Science

A photonic network provides a switched end-to-end data connection by means of a light path.  However, the control of the network must remain in the electronic domain, as photonic memory and logic cannot yet compete with electronics in cost, complexity or power consumption. 

The star network topology has a number of advantages over mesh or ring networks for all-photonic networks.  It can be shown to be equivalent to a single stage switch, and can therefore be non-blocking.  Although it is a simple network there are a number of design challenges for a geographically large network in the coordination of switching functions and engineering.

In this talk we will cover the development of techniques to optimize photonic star networks starting with wavelength-switched networks and finishing with the photonic timeslot-switched Agile All-Photonic Network.  In developing these techniques we have to respect the limitations of photonic technology as well as the opportunities it provides. 

Toward a Reliable Data Transport Architecture for Optical Burst-Switched Networks

Vinod Vokkarane

University of Massachusetts, Darmouth

Computer and Information Science

An Optimization Framework using Lagrangean Relaxation and Subgradient Method  

Oliver Yang and Yiming Zhang  

School of Information Technology and Engineering (SITE)

University of Ottawa

Route Optimization Techniques for Multi-layer Planning Tools

Qinglin Wang and Scott Parker

Nortel

The goal of optimal routing in multi-layer planning is to route demands/wavelengths in the way that the required network resources are minimized with the resultant routes satisfying the routing restrictions of different traffic classes (i.e., length/delay, protection scheme) .

Critical design considerations over multi-layer networks include efficiently homing and packing demands into logical routing entities, logical circuit expressing, minimum-cost routing and efficient protection bandwidth sharing. In physical layer, wavelength regeneration is minimized to reduce OEO components, and optimal  regenerator placement and wavelength assignment are required to minimize the number of required transmission systems. The various optimization techniques required to achieve the design goals will be discussed in this workshop.

Reliable Topological Design of Agile All Photonic Networks

Ning Zhao, Anton Vinokurov and  Lorne Mason

McGill University

In this presentation, the methods and tools for the design and analysis of an Agile All-Photonic Network (AAPN) will be introduced. The layered topology is comprised of edge nodes, selector switches and optical core space switches. The problem of network topological design and dimensioning will be discussed. Also the reliable connectivity and multi-layer restoration strategies is studied in our research. The software tool named Topological Design Tool (TDT) to display network infrastructures and evaluate their performance will also be demonstrated.