Future Optical Networks - OSA Publishing

4684

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 24, NO. 12, DECEMBER 2006

Future Optical Networks Michael J. O’Mahony, Senior Member, IEEE, Christina Politi, Student Member, IEEE, Dimitrios Klonidis, Member, IEEE, Reza Nejabati, Member, IEEE, and Dimitra Simeonidou, Member, IEEE

Invited Paper

Abstract—This paper presents views on the future of optical networking. A historical look at the emergence of optical networking is first taken, followed by a discussion on the drivers pushing for a new and pervasive network, which is based on photonics and can satisfy the needs of a broadening base of residential, business, and scientific users. Regional plans and targets for optical networking are reviewed to understand which current approaches are judged important. Today, two thrusts are driving separate optical network infrastructure models, namely 1) the need by nations to provide a ubiquitous network infrastructure to support all the future services and telecommunication needs of residential and business users and 2) increasing demands by the scientific community for networks to support their requirements with respect to large-scale data transport and processing. This paper discusses these network models together with the key enabling technologies currently being considered for future implementation, including optical circuit, burst and packet switching, and optical code-division multiplexing. Critical subsystem functionalities are also reviewed. The discussion considers how these separate models might eventually merge to form a global optical network infrastructure. Index Terms—Optical communications, optical networks.

I. I NTRODUCTION —E MERGENCE OF O PTICAL N ETWORKING

T

HE INVENTION of the laser by Schawlow and Townes in 1958, followed by the work of Kao and Hockham on optical fibers in 1965, and the subsequent demonstration of optical fiber as a practical communication medium by Maurer et al. in 1970 brought into being a technology platform capable of supporting national and global communication requirements for the 21st century and beyond. In the late 1970s, fiber began to replace coaxial cable as the transmission medium in the trunk systems of telecommunication networks, bringing many advantages both technical and economic. The creation of the Internet (with Transmission Control Protocol (TCP)/IP) in 1983 and subsequently the World Wide Web in 1993 sparked the growth of data traffic on the network, and

Manuscript received April 19, 2006; revised September 20, 2006. M. J. O’Mahony, R. Nejabati, and D. Simeonidou are with the Photonic Networks Research Laboratory, University of Essex, CO4 3SQ Colchester, U.K. (e-mail: [email protected]; [email protected]; [email protected]). C. Politi was with the Photonic Networks Research Laboratory, University of Essex, CO4 3SQ Colchester, U.K. She is now with the National Technical University of Athens, 10682 Athens, Greece (e-mail: [email protected]). D. Klonidis was with the Photonic Networks Research Laboratory, University of Essex, CO4 3SQ Colchester, U.K. He is now with the Athens Information Technology Centre, 19002 Athens, Greece (e-mail: [email protected]). Color versions of Figs. 2–5 are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JLT.2006.885765

Fig. 1.

Evolution of transport and service bit rates.

in 2002, or thereabouts, the amount of network data traffic exceeded that of voice traffic. In the decade from 1985 to 1995, four significant events heralded the possibility of optical networking, where both transmission and switching might be based on optics. These were 1) the realization of optical amplifiers allowing 2) the economic deployment of wavelength division multiplexing (WDM), 3) the demonstration of an optical cross-connect (OXC) enabling the rapid reconfiguration of lightpaths based on wavelength channels, and 4) the convergence of service and transport transmission rate. Early visions of optical networking considered, for example, the deployment of OXCs to form an extension to the existing synchronous digital hierarchy/synchronous optical network (SDH/SONET) network layers—the optical layer. In this paper, an OXC is defined as a general wavelength switch that can be realized in an all-optical (transparent) manner [optical input, optical switch fabric, optical output (OOO)] or in an opaque manner [optical input, electrical switch fabric, optical output (OEO)] through choice of technologies. The optical layer would enable long-haul transit traffic to bypass the main switch nodes and hence reduce the size and cost of the digital cross-connects (DXCs). Demonstrations of such reconfigurable networks were carried out in Europe and the United States in 1994 [1], [2]. The convergence of service and transport wavelength bit rates around the year 2000 (Fig. 1), at a bit rate of 10 Gb/s, opened the possibility of direct interfacing between, for example, an IP network and an optical transport network employing WDM and OXCs, where the granularity of the network directly matched the router interface rate. The figure also shows that convergence at 40 Gb/s occurred in 2005 with the availability of 40-Gb/s routers and engineered 40-Gb/s dense WDM (DWDM)

0733-8724/$20.00 © 2006 IEEE

O’MAHONY et al.: FUTURE OPTICAL NETWORKS

4685

TABLE I RESIDENTIAL SERVICE REQUIREMENTS

Fig. 2.

Schematic of telecommunications network.

transmission systems [3]. Each wavelength, of course, supports many traffic streams. Fig. 2 shows a schematic of the possible future telecommunications network showing core, metro, and access layers, which builds on earlier concepts and will be used to discuss views of future optical networks. The core cloud represents an optical network comprising a number of nodes interconnected by amplified fiber links employing DWDM. The nodes comprise an OXC in conjunction with a network service element (SE). The OXC supports the bypass function and allows specific wavelength channels to be dropped to the SE, which, for example, could be an IP/multiprotocol label switching (IP/MPLS) router, an SDH/SONET DXC, or an optical burst, packet, or Ethernet switch, as discussed later. At the network edge, traffic is mapped onto the network services via an edge interface/router, which can perform either user network interface (UNI) or network node interface (NNI) functionality as defined by the optical internetworking forum [4], depending on what is connected to the core. A control plane is required to establish paths across the data plane as requested at the network edge, mapping, for example, an IP/MPLS stream from the metropolitan area network onto a specific wavelength; path establishment can be done in a centralized or distributed manner. Fig. 2, however, shows only one of a number of network models, based on optical technology, currently being developed to satisfy an increasing diversity of users with greatly differing service requirements; networks to support scientific users are another example and are discussed below. Currently, operator business models do not encourage the development of a homogeneous network, and so, interoperability issues are arising between these different (heterogeneous) networks. This paper addresses two developing optical network models, representing residential/business (telco) and scientific users, to understand their requirements and evolution paths. The key (optical layer) functionalities required in these networks are considered (rather than the detail of all the technology issues that inevitably are part of any evolution) with a view to understanding how a future interoperable global heterogeneous optical network with end-to-end connectivity and possibly eventually transparent (homogeneous) networking might be achieved. It is recognized that a number of functions, for example, dispersion compensation, are currently migrating to the electronic domain, but these topics are not considered here. As discussed earlier,

optical networking includes the concept of OXCs (wavelength switches) using either an OEO switching or OOO approach; the latter is currently a more futuristic approach supporting network transparency, which may well solve many problems including cost issues related to the reduction in the required number of OEO interfaces. However, recent approaches by Infinera [5], where all the interfaces are integrated on a chip, challenge this simple viewpoint. In this paper, Section II looks at applications and their requirements, Section III considers the importance of future networks together with regional plans for optical infrastructures, Sections IV and V look at two emerging network models as represented by national telecommunication networks and national research and educational networks (NRENs), Sections VI and VII consider key networking approaches and technologies, Section VII looks at a possible future global heterogeneous optical network serving all users, and Section VIII summarizes. Here, it should be noted that the views in this paper are essentially research led, and the adoption of many of the approaches and technologies highlighted are, as always, subject to commercial considerations that can change rapidly, as seen in the year 2000. II. 21 ST C ENTURY : E DGE OF THE D ATA T SUNAMI Between the years 2000 and 2003, the volume of data grew from 3 billion to 24 billion gigabytes, with 93% of all data being born digitally [6]. This highlights the start of a huge wave of data to be expected over the next five to ten years as many traditional services and industries move from analog to digital (e.g., TV broadcasting and movie making), and the spread and development of e-services across government, health, and security become more available and acceptable. In addition to the increased expectations from residential and business users, new requirements from scientific users are driving the deployment of high-performance optical networking. Examples of these drivers are as follows. A. Residential Users The bandwidth requirements for fixed home users have been estimated at 100 Mb/s (downstream) in the near future, as shown in Table I [7], the main demand arising from “Triple

4686


Play” services (already launched in a number of EU countries) that include voice data and video, and mobile users will also require high bandwidth access, estimated as 30 Mb/s in this model. These figures would be greatly increased if, as expected in Japan, superhigh-definition TV would also become a broadcast service, moving access speed requirements into the gigabits per second regime. B. Large Business/Enterprise Users Large business/enterprise users require access to high symmetric bandwidths (e.g., up to 10 Gb/s) for virtual private networks, disaster recovery, storage, etc. C. Scientiﬁc Users (Currently Served by NRENs) Application examples include the following. 1) High-Energy Particle Physics: The next generation of experiments at the Large Hadron Collider in CERN will produce data sets measured in tens of petabytes per year that can only be processed and analyzed by globally distributed computing resources. Experiments require deterministic transport of 10- to 100-TB data sets, and a 100-TB data set requires a throughput of 10 Gb/s for delivery within 24 h. Thus, optical network services will be crucial to this discipline where dedicated and guaranteed bandwidth is required for periods of days. 2) Very Long Baseline Interferometry (VLBI): VLBI is used by radio astronomers to obtain detailed images of cosmic radio sources, where the combination of signals from two or more widely separated radio telescopes can effectively create an instrument with a resolving power proportional to their spatial separation. e-VLBI [8] will use high-speed networks to transfer telescope data to a correlator, and the availability of optical network services at multi-gigabits per second (10–40 Gb/s) throughput will greatly increase the capability. 3) e-Health: Remote mammography poses challenges for the deployment of supporting IT systems due to both the size and the quantity of images, with networks required to transport 1.2 GB of data every 30 s. The availability of optical network services offering real-time guarantees is important in this field. All these applications, whether residential, scientific, or business related, look for high bit rate access, wavelength and subwavelength granularity, and quality-of-service (QoS) guarantees; these qualities can be delivered through optical networking. III. M OVE T OWARD P ERVASIVE AND U BIQUITOUS N ETWORKS —R EGIONAL P LANS In the face of this wave of new data-oriented services, the traditional telecommunications network is seen as transforming to a generic and data-centric communications network optimized for data. Most regions of the world have a vision of moving to a situation that in Japan is called “ubiquitous network society” and in Europe is called “ambient intelligence.” By this is meant an environment (comprising wired and wireless network) where one can communicate effortlessly and access key information resources in a straightforward manner. To achieve these ideals, the ultimate network will have as its main building blocks

a fixed optical network platform accessed through wireless (Wi-Fi, WiMax, UMTS) and wired infrastructures, such as fiber to the premises (FTTP). Moving to such an all-pervasive networked society, however, puts increased demands on network reliability and security; it must be there when it is needed and have the ability and flexibility to interface and integrate multiple technologies and service requirements. The importance of future national network infrastructures is mirrored in the ongoing discussions on how a new Internet might be realized. It has been commented [9] that the Internet is expanding from an “information service” to a “critical infrastructure” for all aspects of society, and so, new network architectures must evolve to overcome many of the problems inherent in the current Internet. NRENs and experimental network testbeds are currently being used (in part) to understand how a new Internet might be constructed. The inference is that at all levels, communication networks are no longer just desirable but critical for a nation’s development and security. Regional views on expected growth in capacity demands, traffic profiles of new services, and the need to move to datacentric networking are reflected in the scope of the major research programs and their associated projects funded by regional governments. In Europe, the European Commission (EC) periodically releases tranches of funding for R&D. Currently, there are no overarching roadmaps for future networks, but a number of very large “integrated projects” identify their own vision of the future. The NOBEL project [10], for example, studies the evolution of core and metropolitan optical transport networks, supporting end-to-end QoS, with intelligent data-centric solutions based on automatic switched optical network (ASON) and generalized MPLS (GMPLS) [3] together with optical burst switching (OBS) and optical packet switching (OPS). The investigation of OPS has had long-term support within EU and national programs with projects such as ATMOS, KEOPS, DAVID [11], and OPSNET/OPORON [12], [13] developing the technology and its application over a period of about 15 years; however, it is still seen as a very future technology. On the other hand, OBS and GMPLS approaches are viewed as realistic possibilities for more near-term deployment, and research in these areas is also echoed in national funding in many countries. Recently, BT (U.K.) has committed itself to moving to a converged national network solution (BT 21CN Network) with an IP/MPLS core [14]. This change is also under way in other countries (Netherlands and Australia) and is significant in that the architecture opens the path to a full optical transport network at some point in the future, with the possible replacement of OEO switches and regenerators with OOO technologies. Europe is also characterized by a large number of NRENs that exist in most member countries. These networks are nationally funded and commonly used for research into scientific applications (of the type discussed above). The EC funds an overlay network GEANT [15] that provides interconnections between these national networks and provides international connections, for example, to the United States. These networks look to lambda networking based on scheduled or dynamic provision of lightpaths.


The Pacific Rim countries of Japan and Korea are well known for their ambitious plans in relation to broadband deployment and enhanced network infrastructures. Japan has well-defined R&D programs; some are initiated by the Japanese government with a view to evolving a legacy telecommunications network to an IP over WDM network. Such programs move in tandem with operator plans; for example, NTT plans to migrate 30 million customers to FTTP and IP telephony by 2010. The most recent program focused on developing an all-optical transport network with terabit capability [16]. This paper comprised the study and development of photonic nodes such as fast OXC based on microelectromechanical system (MEMS) together with control plane (such as GMPLS), OBS, and high-bit-rate ultralong transmission based on DWDM (up to 1000 channels) and optical time division multiplexing (OTDM). For photonic routing, targets are related to the feasibility of 100-Tb/s routers and network architectures appropriate to a terabit-class wavelength-routed optical network. The current program seeks to understand how this optical platform can support new bandwidth demanding applications such as grid computing and real-time applications like video, digital cinema, and network storage. It is anticipated that broadcasting and streaming of high-quality video using high- or superhigh-definition TV (4.5 Gb/s) together with grid computing and e-science applications may need to be supported, many of these on FTTP. Key targets include 160-Gb/s multilevel transmission systems using differential quadrature phase shift keying/quadratic-amplitude modulation, etc., aiming at bandwidth utilization of more than 2 bits/Hz by 2010. Ultrafast signal processing functions such as optical gating, optical-3R, OTDM, optical multiplexing/demultiplexing, and optical wavelength conversion are seen as important. South Korea is known to (currently) have the world’s highest (> 60%) penetration of broadband (> 4 Mb/s) Internet access. The broadband convergence network (BcG) is the Korean vision of the ITU-T next-generation (NG) network defining future transport and access architectures and is expected to support 20 million users by the year 2010. In the access network, technologies will be FTTP with E-passive optical network (E-PON) and WDM-PON. The BcG transport network will be a managed optical network with support for end-to-end QoS provisioning. BcG is seen as a ubiquitous network that will also support grid applications. China has a number of optical network testbeds for ASON and OBS, which investigate the support of future IP-based services. Moreover, there is a strong interest in optical grid networking with associated research into architectures, control plane functions, and interfaces [21]. In the USA, funding for research (nonindustrial) is through the National Science Foundation and for large projects often through DARPA. Current major photonic activities include studies on optical code-division multiplexing (OCDM) and terabit router technology; the latter represented by projects IRIS [17] and LASOR [18] whose goal is to realize 100-Tb/s routers. Both projects take the route of OPS, which offers attractions in terms of footprint and power requirements. The strong interest in optical networking in the USA is reflected in the existence of a number of national testbeds,

4687

Fig. 3. Network evolution.

which enable the interconnection of scientific users and support research into future networks. For example, the National Lambda Rail (NLR) [19] is a high-speed national computer network that is also used as a network testbed for experimentation with NG large-scale networks. Links in the network use DWDM at 10-Gb/s/channel. NLR’s services are already in use by many network research projects, for example, the NSF OptIPuter project and Internet 2’s Hybrid Optical Packet Infrastructure (HOPI) project, which looks at a future infrastructure comprising an IP core network together with an optically switched wavelength set, for the dynamic provisioning of high-capacity paths. Currently being proposed is a new national facility GENI [20], which includes a global experimental facility designed to explore new network architectures with the broad scope of understanding new paradigms for Internettype networks. IV. E VOLUTION T OWARD N ATIONAL O PTICAL T ELECOMMUNICATION N ETWORKS Fig. 3 outlines a possible evolution route for the network structures and technologies that may appear in the future optical (transport) network; a version of many such diagrams have been presented over the years. Progress toward optical networking has been much slower than envisaged in the late 1990s, and currently, the first real steps toward networking are seen in the deployment of reconfigurable optical add–drop multiplexers (ROADMs), in particular those based on a multiport wavelength selective switch (WSS) [22]. These devices have the functionality of OXCs—an exciting development. In the following discussion, the evolution is examined from three viewpoints. A. Transmission Speed Current networks employ amplified DWDM systems with individual channel bit rates up to 10 Gb/s to connect main switching centers, and the industry is on the verge of deploying 40-Gb/s systems. There are many reasons for believing that bit rates will increase beyond 10 Gb/s and perhaps even to 160 Gb/s

4688

(Fig. 3). For example, 1) in the past, it has always been advantageous to move to high bit rates from a cost viewpoint; 2) although optical amplifiers are reasonably bit rate agnostic, most of the more advanced functions currently considered for future networks such as all-optical regeneration, wavelength conversion, dispersion compensators, etc., can only operate on a single wavelength; and 3) when future OOO switches are deployed, then increasing the bit rate can help in reducing the port count required of the switch fabrics [23], as large OOO switch fabrics are difficult to realize, for example, in the case of an OPS where it is difficult to scale fabrics; for wavelength switching, waveband approaches [24] can be used to mitigate these technological issues. Increasing the bit rate, however, leads to a more demanding requirement for system design to minimize the effects of dispersion (chromatic and polarization) and nonlinear effects; thus, there is the increasing interest in modulation techniques such as differential phase-shift keying (DPSK), which is more robust to transmission impairments than intensity modulation. B. Network Switching Much research has focused on moving from the traditional circuit-based switching to a more dynamic and data-centric network enabling rapid lightpath reconfiguration and providing subwavelength granularity, as needed by the new applications discussed above. In its present form, the network has a complex layering to allow the simultaneous support of data and voice services; in this mode, data (IP) may be encapsulated into asynchronous transfer mode (ATM) cells (or SDH/SONET frames) for transmission across the point-to-point connections between DXCs. Currently, the network is changing to a more data-oriented version of SDH/SONET (NG SDH/SONET). Fig. 3 illustrates the possible further stages in the network switching evolution. Fig. 3(a) represents the move to a more data-centric and dynamic switching model using an ASON [4] architecture, which would allow automated lighpath provisioning and supports NG-SDH/SONET with DXC or OXC (OEO) switching in the data plane. Fig. 3(a) also shows that a move to IP/MPLS (i.e., IP routers) or GMPLS (with OEO or OOO wavelength switches) architectures is foreseen, which provides an enhanced dynamic capability GMPLS to allow all transport modes, circuits, burst, and packets to be supported and can be deployed in either a centralized or a distributed mode. This represents one of the options to build a “converged network,” where the backbone is a multitechnology IP/GMPLS/ OEO/OOO network supporting all services (voice, data, video), which may overtake the ASON architecture. It is also the case that in recent times carrier Ethernet [25] (based on native Ethernet or MPLS) looks increasingly attractive across all layers of the network. Indeed, within the U.K., some (small) network providers already operate national converged networks with Ethernet switching elements. The move toward 100 GbE standards illustrates the importance of this technology and hints of future major roles in NG networks. Fig. 3(b) represents a move to a user-centric design based on OBS with GMPLS (OBS/GMPLS); this technology provides subwavelength granularity and is also of interest to future optical grid


networks [26]. Finally, Fig. 3(c) represents the move to an OPS network, where MPLS provides a common (across electrical/optical domains) control plane. OPS offers the finest granularity and is still seen as the ultimate switching technique, but its success will depend on many technological advances; these options are discussed in more detail in Section VI. The metro/aggregation network (Fig. 2) delivers data from the access to the edge device for processing and will likely use DWDM supporting, for example, carrier Ethernet, but it is the access network that provides the key to the future, in particular, the move toward FTTP. C. Access PONs FTTP supports the PON concept, which has been studied for over 20 years but now increasingly deployed. PONs offer multiservice (voice, data, video, and telemetry) and multiprotocol (IP, TDM, and ATM) support and thus are a very flexible infrastructure. A number of possibilities have been studied, which exploit the basic PON concept but with a view to reduce costs. Two examples follow. Long-Reach PONs: Some operators [27] see the possibility of merging metro and access in a long-reach (amplified) PON. Optical amplification is included to boost the power budget and increase bandwidth, range, and number of splits. Long reach is used to bypass the metro network and terminate at a core edge node; this enables the removal of the local exchange or remote concentrator site. In the U.K., this would require 100 of these core edge nodes with long reach spans of 100 km and bit rates of at least 10 Gb/s. CWDM/OCDMA PONs: Coarse WDM (CWDM) allows low-cost WDM to be deployed (as device cooling is not required), and CWDM uses a channel spacing of 20 nm, so only eight or so channels can normally be deployed. Optical codedivision multiple access (OCDMA) has been demonstrated [28], [29] as a robust complementary technology that could also be deployed in conjunction with CWDM to increase the number of users. As with WDM, OCDMA offers the possibility of translation from one channel to another (code translation), opening many interesting possibilities.

V. E MERGENCE OF AN A LTERNATIVE N ETWORK M ODEL Following the realization that a computer bus speed cannot match a lambda-based optical network (at 10 Gb/s), it was suggested that it should be possible to create a tightly integrated cluster of computational, storage, visualization, and instrumentation resources linked over parallel dedicated optical networks across campus, metro, national, and international scales to support scientific tasks, and this concept is now being demonstrated in network testbeds and NRENs. Scientific and grid applications have complex workflows, which will make extensive use of an optical network, for example, to transfer huge amounts of data between storage and computing or visualization resources. In many ways, NRENs mirror the possible future requirements on telecommunication networks and may evolve to use a similar set of optical technologies (e.g., OBS).


Fig. 4.

NRENs.

NRENs have high-speed optical backbones and can already offer dedicated channels (wavelengths) to individual research users. Switching within these networks is, at present, generally achieved through OEO switches, allowing wavelength routing. NRENs are seen as an important vehicle for advancing innovation and discovery and as a platform for undertaking network research. In Europe, GEANT connects 30 such NRENs (with bit rates up to 10 Gb/s); in Japan, JGN2 provides an research and development environment (at 10 Gb/s); Abiline in the United States interconnects 50 states; Canada has CA∗ ,net; and China’s CERNET2 connects 20 cities (at 10 Gb/s). Global Lambda Integrated Facility (GLIF) [30] is an international virtual organisation whose members are NRENs and promotes lambda networking for international collaboration and research. In today’s NRENs, resources are provided either exclusively for an application or are shared on a best-effort base. Many emerging applications demonstrate a dynamic nature and require connectivity, bandwidth, or QoS that may change with time and/or user profile. The administrative and technical overhead to provide such network services and the high associated costs currently inhibit a wider commercial adoption of such demanding applications. A short- to medium-term solution to the above requirements is the ability to negotiate flexible service level agreements between the application and the resource provided. This is a subject of current research and standardization within the Global Grid Forum [31]. As a longer-term solution, a new network service model that allows users or/and applications to adapt the network to their applications is discussed here. This network model will require the network topology to migrate from the traditional edge-core telecom model (Fig. 2) to a more distributed model. In this type of network, the user may have the ability to control routing in an end-to-end manner and set up and tear-down lightpaths between routing domains. To facilitate this level of user control, users or applications will be offered management/control of the network resources (i.e., bandwidth allocation at the wavelength and subwavelength level). These resources could be leased and exchanged between users. Such new topological solutions will have a direct impact on network management and control as well as the network infrastructure, including network elements and user interfaces to enable and support these functionalities. An important step toward this is the ability to make dynamic resource reservations immediately when needed or in advance for a time period in the future. Dynamic optical networking can satisfy the bandwidth and reconfiguration requirements of this model together with software tools and frameworks for

4689

end-to-end on-demand provisioning of network services across multiple administrative and network technology domains. Fig. 4 illustrates the global interconnection of a number of NRENs using interconnecting networks such as GEANT or GLIF. Within each network, the choice of switching technology and transport format is an important consideration that influences the ability to deliver NG network services. As discussed above, optical switching offers bandwidth manipulation at the wavelength and subwavelength level (e.g., with optical circuit/burst/packet switching) as well as the capability to accommodate a wide variety of traffic characteristics and distributions. The transport format determines how signaling and control messages as well as data are sent from the user/client to the optical network and depends on the form of switching. Thus, for 1) circuit switching, signaling is sent in conjunction with the data or over a dedicated wavelength or SDH/SONET connection, and for 2) OBS/OPS, signaling is sent using a signaling packet or control burst; hybrid approaches can also be used. The service model proposed here does not depend on transport technology (i.e., CS, OBS, OPS). Actually, the first implementation will address wavelength-switched optical infrastructures. However, subwavelength granularity will enable the offering of dynamic network services to a wider range of users and applications. The choice of transport format is mainly driven by an understanding of the traffic characteristics generated by users and their applications. For example, wavelength switching may be the preferred solution for moving terabytes of data from A to B but is inappropriate for video games applications. The diagram also shows other important elements of this global network. Each NREN has an associated network resources provisioning system (NRPS) currently being developed and deployed around the world by different international organizations. These systems are based on the abstraction of network resources. For instance, the user-controlled lightpath provisioning system (UCLP) can be thought of as a configuration and partition manager that exposes each lightpath in a physical network and each network element associated with a lightpath as an “object” or “service” that can be put under the control of different network users to create their own IP network topologies. UCLP, as proposed by CANARIE, is an NRPS that deals with the abstraction of network resources as objects to allow end users to manage them in order to build reconfigurable discipline or application-specific networks. More specifically, UCLP [32] is a set of distributed services that are used to establish and tear down end-to-end connections across an optical network. The service plane mainly involves application-level middleware and Application Programming Interfaces (APIs). The service middleware in this architecture will provide a unified application execution environment relying on a high-bandwidth QoS-enabled network infrastructure of global scale. The middleware could implement service abstractions exposing network resources (end-to-end QoS management). Hence, it will be possible for an application to request network constraints (e.g., guaranteed bandwidth or latency among computational nodes).

4690


VI. K EY N ETWORK T ECHNOLOGIES Research on advanced switching and networking technologies is mainly determined by the needs identified in the core of the network, where large amounts of data with different characteristics (generated by different services) require efficient switching in terms of bandwidth utilization, resource management, and performance. Following the switching technology evolution plan presented earlier (Fig. 3), research in the area of optical networking can be categorized in a) the near term as automatically reconfigurable circuit-switching solutions, b) near to long term as OBS solutions, and c) long term as OPS solutions. The feasibility of implementing these solutions is reflected in the results of research studies and demonstrations; these are briefly discussed here to understand feasibility. A. Optical Circuit Switching The widely deployed circuit-switching technology is currently moving toward the deployment of fast and automatically reconfigurable nodes with switching granularity at the wavelength level. This is now represented in standards by the ASON architecture [4]. To enable dynamic networking, the provisioning of lightpaths is automated. This is achieved with an advanced control plane, which provides the necessary signaling capabilities. The user can request, through the UNI, a new lightpath, and the controllers in the network nodes exchange, through NNIs, the necessary information for the lightpath setup (allocation of wavelengths, etc.). Research in this area is at the level of final product development and is carried out by a number of leading companies presenting applicable solutions for a) node design like ROADMs based on WSSs and for b) control planes for dynamic networking, channel provisioning, and management based on IP/MPLS solutions. Demonstrations of these can both be represented as that in Fig. 2, where for IP/MPLS the core has an MPLS router as the SE together with an optical switch (initially OEO, but with possible upgrading to OOO) at the nodes. For the ASON network, the SE represents an OXC controller, and lightpaths are being set up by the control layer. As currently viewed, the ASON network, which is based on SDH/SONET, has advantages in terms of QoS, management, and security, and is able to reduce the load taken from the IP/MPLS network. However, due to the domination of IPcentric traffic, future solutions are focusing on bursty networking models able to handle dynamic segments instead of continuous data. B. OBS Segmented data transmission has been derived by the Internet paradigm in order to mainly offer increased bandwidth utilization and reduced overhead. One idea is to set-up and tear down a path dynamically (similar to circuit switching) but for a much shorter duration, equal only to the duration required to transmit a complete set of data, a “data burst.” This is the fundamental premise of OBS, which is based on the separation of the control and data planes, and the segregation of functionality within the appropriate domain (electronic or optical). Prior to data burst transmission, a burst control packet (BCP)

is created and sent toward the destination by an OBS ingress node (edge router). The BCP is typically sent out-of-band over a separate signaling wavelength and processed at intermediate OBS routers. It informs each node of the impending data burst and sets up an optical path for its corresponding data burst. Data bursts remain in the optical plane end-to-end and are typically not buffered as they transit the network core. The bursts’ content, protocol, bit rate, modulation format, and encoding are completely transparent to the intermediate routers. The main advantages of the OBS in comparison to the other optical networking schemes are that unlike optical wavelengthswitched networks, the optical bandwidth is reserved only for the duration of the burst, and that unlike the OPS network, it can be bufferless, but it also needs a switch reconfiguration speed in the order of microseconds. Due to these implementation issues (easier processing, bufferless operation, and slow switching requirements), OBS is seen as a practical solution and is considered as the next evolution step in future optical networking. The feasibility of OBS technology can be identified by a number of reported results from field trials and testbeds. The most complete demonstrator was set up in Japan under the “OBS network” project and accommodated six nodes with MEMS-based switches, achieving switching times of 1 ms for bursts with a minimum size of 100 ms [33]. In China, a collaborative work between two universities under the TBOBS project resulted in an OBS field experiment that comprises three edges and one core node [34]. In this demonstrator, switching is achieved in nanosecond times utilizing a tunable WSS based on semiconductor optical amplifiers (SOAs). Additionally, high-speed electronics are used for the header control mechanism. The fast switching time allows routing of relatively small packets, 720 µs, transmitted at 1.25 Gb/s. OBS makes an attractive proposition for deployment in metro/wide area networks. Until now, the only way to build networks with more than 10 Gb of bandwidth has required the use of DWDM technology to create point-to-point circuits for every path across a network. This has proven to be both expensive and cumbersome to manage. OBS offers a new alternative by using burst transponders that can communicate directly with multiple destinations across a network. This means no circuit needs to be preprovisioned, and high-speed transponders need not be dedicated for every single communication path. This frees up capital, simplifies network design, and enables the creation of packet metro aggregation networks where bandwidth shifts in real time to where it is needed in the network. A new generation of startup companies [35] is appearing and, for the first time, delivering commercial equipment merging the best of both worlds—the efficiencies of Ethernet packet switching with the bandwidth of DWDM optical technology. OBS has also been identified as a compatible solution for the physical layer infrastructure in grid computing applications with possible realization on NRENs [36].

C. OPS OPS is a purely connectionless networking solution that is fully compatible with IP-centric data traffic and offers the


finest network granularity and optimum bandwidth utilization. In OPS networking, incoming IP packets are assembled at the network edge to form the payload of an optical packet, which has a leading optical header. At the switch node, the header is interrogated and used to configure the associated optical switch for forwarding of the payload. OPS requires the use of more demanding subsystems (than OBS) with intrinsic intelligence to realize adequate packet processing and routing on the fly. The main challenges in OPS are the implementation of the optical header processing mechanism, the development of an intelligent switch controller, the realization of ultrafast switching in nanosecond timescale, and the exploitation of buffering mechanisms to reduce packet blocking. In contrast to OBS, complete OPS demonstrators are mainly restricted to the development of fully functional but small switching elements that simply show the feasibility of fast packet switching on the physical layer with some extensions to the link layer. In the OPSnet project [12], dynamic switching of asynchronous optical packets at 40 Gb/s has been demonstrated in a fully controllable set up able to identify and process the header and route the payload accordingly. In [37], contention resolution on the wavelength level is also considered on a 10-Gb/s packet switching node. Newly developed schemes [38] are based on the same concept and use more advanced electronics for faster clock extraction and processing, while the integration capabilities with the WSS are investigated. A more feasible approach toward the implementation of OPS considers the use of synchronously (slotted) transmitted packets with fixed lengths, but in this case, the hardware overhead is on the implementation of the packet synchronizer at the input. However, slotted solutions are attractive for other applications like computer interconnects. Despite their feasibility limitations, OPS demonstrators assisted the development of numerous ultrafast switching and processing techniques regarding wavelength conversion, header encoding/decoding and processing, label swapping, fast clock extraction, regeneration, and optical contention resolution. Additionally, various switch architectural designs and control protocols have been proposed, which in combination with the significant technological advances over the last years indicated the possible deployment of OPS in the future. However, its future relies on advances in photonic integration that will enable cost-effective subsystems to be constructed.

D. OCDMA OCDMA has been studied as an alternative networking solution able to increase passively the number of users per wavelength. The other solution is OTDM, but this requires active processing. The advantages of OCDMA have been evident for some time through its successful use in wireless networks. For optical networking, its potential for enhanced security, decentralized control, and flexibility in bandwidth granularity provides interesting possibilities to solve these well-known issues in the development of future networks. Additionally, the feasibility of OCDMA has been assisted by newly developed components able to provide simple ways of coding and decod-

4691

ing signals in a passive manner, which is a particularly attractive and cost-effective feature. The principle of OCDMA relies on applying unique orthogonal codes to each user while they share the same bandwidth (wavelength). User codes are multiplexed on to the fiber channel and are matched at the receiver against copies of the codes through an autocorrelation process. The code length is chosen to try and maximize discrimination in the detection process, where multiple access interference (MAI) is a key problem. OCDMA must support many users, and hence, the codes chosen must satisfy certain correlation conditions. Both the time-shifted autocorrelation function of each code and the cross-correlation function between codes must be low when compared to the peak autocorrelation value. The recent development of coherent OCDMA en/decoders allowed the efficient separation of a large number of simultaneously transmitted users providing a feasible solution for low-cost applications in multiuser LAN environments. Also, the combination of OCDMA with CWDM technology can boost the total number of supported users, showing compatibility with PON architectures for FTTP solutions. The combination of OCDMA and WDM technology in a LAN/access environment has been recently demonstrated in [28]. The paper reported the field trial of a 3-WDM × 10-OCDMA × 10.71-Gb/s system over a 111-km field trial. Key aspects of the approach were 1) the use of a multiport encoder/decoder in the central office, which can give multiple optical codes in multiple-wavelength bands (this device gives good correlation properties to suppress MAI and beat noise), and 2) the use of a super structured fiber Bragg grating (SSFBG) (or tunable transversal filter) at the optical network unit (ONU). The use of DPSK–OCDMA with balanced detection is seen as a key enabler over conventional ON/OFF-keying OCDMA with superior noise performance. In recent times, other interesting possibilities for OCDMA have been considered. One such is the concept of code translation, where, analogous to wavelength conversion, codes can be transformed to match users in other subnets (i.e., provide routing). Experiments [29] have also shown how a narrowband spectral-phase-encoded (NB-SPE) OCDMA compatible with existing WDM networks can be used with passive code translation to play a role in ring- and star-based network architectures. Finally, the use of optical codes under a different concept has shown the feasibility to implement an OPS node [39]. Here, optical codes are used as labels in order to distinguish the different headers of the transmitted packets. After header matching, the autocorrelation peak triggers an electronic controller that shows the output port where the packet should be routed. VII. K EY S UBSYSTEMS AND T ECHNOLOGIES The roadmap in Fig. 3 and the above discussion broadly outline a much argued route forward to the future with switching or transmission aspects relevant to a future global heterogeneous optical network. Generally, the picture is that transmission speeds are still predicted to increase as before (e.g., to 160 Gb/s probably using DPSK), within the context of a more dynamic and granular network, supported by appropriate control plane

4692

(e.g., GMPLS with extensions). Many of the key functional subsystems needed for this future flexible network have already been described with experimental demonstrations (e.g., all-optical regeneration at 40 Gb/s has been widely reported [40], as has wavelength conversion, fast tunable lasers, tunable dispersion compensators, MEMS-based switch fabrics, etc.). Others are still gleams in the eye of the network designer with, as yet, no sure route for realization, for example, optical memory and multiwavelength optical regeneration. However, what has become well recognized is that the future of realizing the full potential of optical networks, at affordable cost, lies in the ability to perform good levels of photonic integration. This has not been an easy route to date, but some interesting examples of future approaches are starting to appear. Generally, the integration of photonic components is not straightforward as the substrate technology for various components differs (e.g., lasers are grown on InP, whereas filters benefit from silicon technology). Existing photonic integration approaches are based on hybrid integration, where individual components are laid down and interconnected on a suitable substrate (e.g., silicon); hybrid integration offers full functionality, lower cost, and shorter development times; or on monolithic integration, where a limited set of functions, realizable on a common material, is combined. Recent advances in the integration of Mach–Zehnder interferometers (MZI) illustrate the potential benefits of integration. In a typical device, a pump signal is split between two interferometer arms, and the phase difference between arms is controlled by the optical signal operating on a nonlinear element, typically an SOA. Integrated MZIs can be used for a variety of functions needed in future systems and cannot be realized by assembling discrete components; current commercial devices can operate at 40 Gb/s and have been extensively used for wavelength conversion at 10 and 40 Gb/s with regenerative capabilities [41]. To maximize speed, push–pull configurations have been suggested where, by applying phase changes in both arms, performance can be significantly enhanced, with up to 160-Gb/s operation reported [42]. This configuration has been used in a number of different signal processing applications such as regeneration, add-drop multiplexing, time slot interchange switching [43] format, and wavelength conversion, and other optical processing functions. 3R regeneration can be achieved by using the input signal to gate-extracted opticalclock pulses in a push–pull configuration. The ability to integrate many interferometers in one substrate makes the device very interesting and versatile, for example, to realize logic functions and bursty receivers [44]. Recently, novel approaches based on silicon photonic integration (Silicon Photonics) have been reported by Intel [45]. Silicon is a well-understood material and the basis of lowcost electronics. It is transparent at infrared wavelengths and so can be used to guide light but cannot emit light. Intel is pioneering a process whereby a laser chip mounted in a silicon external cavity forms a tunable laser, which together with silicon modulators (realized by MZI configurations) and photodetectors (with Ge doping) shows promise as a future low-cost integration strategy. The current performance is at 1 Gb/s, but 10–40 Gb/s is predicted.


There are a number of key functions necessary, at a link and network level, to enable efficient operation of the future optical network. Examples are as follows. A. Optical Switching To fully advance to transparent networking, OOO switches are required at microsecond (for OBS), millisecond (for OXCs), and nanosecond (for OPS) reconfiguration times. 3-D MEMS switches (ms reconfiguration) are now available with port counts to 160 × 160 [46], and the recent interest in the deployment of ROADMs means that the functionality of OOO switches is now available. Fast (in nanoseconds) switch fabrics (of high dimension) are realized through the combination of tunable lasers/wavelength converters together with arrayed waveguides; it is hard to see this situation changing. Nevertheless, low-dimension nanosecond switch modules are available [47]. An example based on the total internal reflection in a compound semiconductor power line carrier [48] forms the core of a commercially deployed OPS system. Recent advances in fabrication have made it possible to fabricate compact microring resonators with radii as small as 2.5 µm and quality factors as high as 10 000 by tightly confining semiconductor waveguides [49]. The large field-enhancement factors obtained in these microcavities can result in three to four orders of magnitude reduction in the required switching powers, while the small dimension of the devices helps reduce the cavity lifetime, and hence the switching times, down to the picoseconds regime. With the microring arranged in either the singlecoupler (all-pass) or double-coupler (add–drop) configuration, the device can be used to perform efficient switching, pulse routing, wavelength conversion using four-wave mixing, as well as logic functions; further research is needed to understand the potential of this approach. B. Optical Monitoring As the network performance increases, it is crucial to understand the state of the physical layer so that appropriate routing and remedial actions can be taken. For example, within a network, a variety of bit rates from 10 to 160 Gb/s may exist, with some routes more appropriate to one bit rate than another from the point of view of optical signal-to-noise ratio, residual dispersion channel power, etc. Information should be available through optical monitoring to inform routing decisions at network nodes; this trend now appears in the area of cross-layer routing and is an important one for research. C. Optical Encryption Security will always be of great importance in networks, and it proves difficult to perform electronic encryption at speeds beyond 10 Gb/s. Thus, there is a strong interest in the investigation of encryption possibilities in the optical medium. Two main techniques have been investigated based on 1) quantum cryptography and 2) chaos cryptography. The first is based on secure key distribution, in which the key can then be used to transmit information securely by the conventional algorithmbased encryption. Quantum cryptography guarantees secure


communications by harnessing the quantum nature of photons sent between users. Any attempt to intercept the photons will disturb their quantum state, and as this quantum quality is an integral part of key generation, the disturbance will be detected. Several experiments have extended the key distribution over an ordinary 100-km fiber [50]; however, the technique is sensitive to noise. Chaos-based encryption has the advantage of allowing very high encryption speeds (> 10 Gb/s). It relies on the synchronization of chaotic systems, for example, a pair of unidirectionally coupled single-mode semiconductor lasers subjected to coherent optical feedback or injection. Synchronization means that the chaotic output of the emitter device can be reproduced by the receiver. Once the two lasers have synchronized, the output light of the emitter can be used to encode a message. An experiment over a commercially installed fiber (2.5 Gb/s) showed the possibility of recovering chaotic transmission after 120 km [51].

D. All-Optical Wavelength Conversion and Regeneration All-optical wavelength conversion and regeneration are desirable (including from a power and footprint viewpoint) for networks operating at speeds of > 40 Gb/s. To achieve conversion, the physical properties of a nonlinear element are used to perform a logic function between the input signal and a pump. The main nonlinear elements used are SOA, an electroabsorption modulator (EAM), fiber, a photonic crystal, and periodically poled LiNbO3 waveguides (PPLNs). SOA-based devices, especially quantum dot, and EAMs have the added advantages of compactness and low-energy requirements to trigger nonlinearities. Fibers have an instantaneous response to pulses but, on the other hand, have limited nonlinearity, even in specially designed photonic crystal fibers; hence, long lengths are required. PPLNs require intermediate lengths, and very fast conversion (40–160 Gb/s) has been demonstrated. Considerable research has been done in the area of singlewavelength subsystems. For example, advanced all-optical regenerative schemes at bit rates beyond 100 Gb/s have been recently proposed in [52] as well as well reported demonstrations at 40 Gb/s [53]. Multiwavelength all-optical regeneration, if feasible, will dramatically decrease the cost of DWDM transmission links. A number of techniques are currently being considered, for example, 1) through the mechanism of selfphase modulation, which in principle enables operation at speeds of > 160 Gb/s, and 2) based on the inhomogeneously broadened gain of self-assembled quantum dots in quantum dot SOAs [54].

E. Optical Memory All-optical buffering through fiber delay lines is an approach that requires complex control, and packets are delayed rather than stored. Recently, in the framework of LASOR [18], integrated delay lines have been developed as it has been shown [55] that a small number of low-depth buffers are sufficient for an OPS network.

4693

Nevertheless, recent research has established that it is possible to exercise control of the velocity of light pulses propagating through a material. Light propagation with a very low group velocity (slow light) has been observed in atomic vapors and solid-state crystal via electromagnetically induced transparency and coherent population oscillation. Recently, direct timedomain measurements in SOA quantum wells showed controllable delay up to 1 ns; this results in a direct measurement of group velocity of < 200 m/s, giving a slow down factor of up to 1.5 × 106 [56]. Various other methods have been proposed and demonstrated, and the area is still one of strong research. Critics of this technique have concluded that there is a delay–bandwidth product that inevitably limits the achievement of reasonable delays at high bit rates. VIII. F UTURE G LOBAL H ETEROGENEOUS O PTICAL N ETWORK In the above discussions, it has been shown that there are ever-growing demands on future networks from a number of directions. The historic telecommunications network is seen as evolving to a fully pervasive or ubiquitous network supporting a wide variety of users. Key to this goal is the evolution to a multiservice platform comprising (eventually) an optical core with wavelength and subwavelength granularities together with an advanced control plane such as GMPLS. The core data plane will support electrical circuit switching (e.g., with DXCs), electronic packet switching using IP/MPLS routers, OBS using microsecond optical switches with appropriate edge interfaces to set incoming data on to specific wavelengths, and finally, possibly OPS using nanosecond optical switches and appropriate edge interfaces. The optical transport core will be accessible to the user via fixed and wireless access networks, providing high bandwidth availability. Single-channel transmission speeds will increase in the very near future to 40 Gb/s (matching router interface rates), and it is likely that development will continue to move to higher rates such as 160 Gb/s, perhaps using DPSK, and via OTDM. Continued development in photonic integration will enable appropriate optical processing subsystems, for example, alloptical regeneration, to be realized supporting these changes. From the scientific user direction, it was shown that dedicated optical networks, for example, NRENs, based on dark fiber and readily available optical technology are already being put in place to satisfy the requirements of these users for a dynamic network with specific requirements on bandwidth, latency, and availability, to access global processing resources, transfer high data volumes with QoS, and provide a secure infrastructure for future high-end services. These networks, with their welldefined requirements, will likely continue to rely on dynamic circuit-based switching, but the advantages of OBS for grid applications make it likely that this technology will move to these networks as well. Fig. 5 shows how these separate networks might eventually coexist to form a global heterogeneous optical network, increasingly supported by advanced optical technology. The figure shows how a core telecommunications network can provide a platform for both residential users through the metro/access

4694

Fig. 5. Global heterogeneous optical network.

route or for scientific communities through the interconnection with NRENs. This solution is one that allows telecom operators to build a ubiquitous network, which also supports the flexible services needed by high-end researchers. The future vision here is of operators providing transport services to research users and legacy ones (i.e., residential/business users), on different sections of the same network infrastructure or sharing the same infrastructure, but under a single unified control and management plane. It is necessary that this common control plane should interwork with the existing NRPS used by NRENs in order to dynamically implement worldwide optical transport services. GMPLS has proved to be the most efficient telecom-oriented solution for the fast and automated provisioning of connections across multitechnology (IP/MPLS, Ethernet, SDH/SONET, DWDM, OBS, etc.) and multidomain networks: GMPLS enables advanced network functionalities for traffic engineering, traffic resilience, automatic resource discovery, and management. However, GMPLS is not natively designed to support the flexible network services described above, and a major technical challenge rises from the interworking of the application platforms and of their resource management systems with the underlying NG optical networks powered with the GMPLS control plane. This network vision needs to ensure a service boundary is kept between the users and the network service provider but also enable differentiated and configurable interfacing procedures. According to this approach, the boundary of the network will be transparent only for “power users” (i.e., scientific and high-performance grid) who will • obtain network topology information to be used by resource brokering entities (i.e., NRPS); • request specific lightpaths by providing lightpath description; • communicate application layer information (e.g., CPU, storage, etc.) into the GMPLS control plane, which will


allow dissemination of this information within the same virtual. The proposed coordination between the telecom network control plane, the NRPS, and the service plane on top of the switched physical layer will enable end-to-end dynamic service provisioning across a global heterogeneous optical network infrastructure. In this scenario, therefore, it is likely that progress will continue toward a common or shared optical infrastructure supporting residential, business, and scientific users. All these users require a more dynamic network with wavelength and subwavelength access, which will likely eventually be supported, for example, through OBS or indeed (in the far distance) OPS. The common core network will continue to move to higher speeds with an eventual move from OEO to OOO switching. Photonic integration is seen as a key to enable high-performance optical networking. For the global heterogeneous optical network, however, there are many challenges on both physical and network layers, requiring much cooperation among traditional electronic engineers, computer scientists, and users. IX. S UMMARY This paper presented a story of the evolution of optical networking from early concepts to its strong progression to provide the main high-capacity and flexible platform necessary to support the near and future requirements of a global network supporting the communication requirements of a growing diversity of residential, enterprise, and scientific users. Many of the switching and subsystem technology solutions necessary to achieve these aims have already been glimpsed but need considerable progress in integration techniques to overcome deployment cost problems. There is also much research to be done at the software level to ensure interoperability between networks at a global level. Eventually, hardware and software evolution may lead to increased levels of transparency in these future optical networks. R EFERENCES [1] J. Zhou, R. Cadeddu, E. Casaccia, C. Cavazzoni, and M. J. O’Mahony, “Crosstalk in multiwavelength optical cross-connect networks,” J. Lightw. Technol., vol. 14, no. 6, pp. 1423–1435, Jun. 1996. [2] R. E. Wagner et al., “Monet: Multiwavelength optical networking,” J. Lightw. Technol., vol. 14, no. 6, pp. 1349–1355, Jun. 1996. [3] T. Freeman, “DWDM platform set for 40 G deployment,” Fibre Syst., vol. 3, no. 1, p. 23, Jan./Feb. 2006. [4] N. Larkin, ASON & GMPLS; The Battle for the Optical Control Plane. [Online]. Available: http://www.dataconnection.com/network/download/ whitepapers/asongmpls.pdf [5] R. Nagarajan, C. Joyner, and R. Schneider, “Large-scale photonic integrated circuits,” IEEE J. Sel. Topics Quantum Electron., vol. 11, no. 1, pp. 50–65, Jan./Feb. 2005. [6] J. S. da Silva, “Broadband R&D in the context of convergence,” presented at the Broadband Europe Conf., Bordeaux, France, Dec. 2005, Paper M01.03. [7] J. J. Lepley, “The evolution of the access network: Options for fibre penetration into the access network,” London, U.K.: MUSE Summer School Broadband Access Technol., Jul. 4–5, 2005. [Online]. Available: http:// www.istmuse.org/Documents/NOC2005/Summer_School/Jason_Lepley_ Fibre_Penetration_in_XDSL_Networks.pdf [8] [Online]. Available: http://www.jb.man.ac.uk/news/evlbi/ [9] B. St. Arnaud, “Future internet issues,” in Proc. OECD ICCP Workshop, Paris, France, Mar. 2006. [Online]. Available: http://www.oecd.org/ dataoecd/43/63/36274169.pdf


[10] C. Cavazzoni, D. Colle, A. Di Giglio, G. Edwall, G. Eilenberger, G. Ferraris, H. Haunstein, S. Herbst, M. Jaeger, G. Lehmann, J. F. Lobo, A. Manzalini, S. Santoni, and M. Schiano, “Evolution of optical transport networks in Europe: The NOBEL project vision,” presented at the Broadband Europe Conf., Bordeaux, France, Dec. 2005. Paper T04A.01. [Online]. Available: www.bbeurope.org [11] A. Stavdas, S. Sygletos, M. O. Mahony, H. L. Lee, C. Matrakidis, and A. Dupas, “IST-DAVID: Concept presentation and physical layer modelling of the metropolitan area network,” J. Lightw. Technol., vol. 21, no. 2, pp. 372–384, Feb. 2003. [12] D. Klonidis, C. T. Politi, R. Nejabati, M. J. O’Mahony, and D. Simeonidou, “OPSnet: Design and demonstration of an asynchronous high speed optical packet switch,” J. Lightw. Technol., vol. 23, no. 10, pp. 2914–2925, Oct. 2005. [13] R. Nejabati, D. Klonidis, G. Zervas, D. Simeonidou, and M. O Mahony, “Demonstration of a complete and fully functional end-to-end asynchronous optical packet switched network,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper OWP5. [14] J. Ryan, “An outlook for optical executatives,” OSA Executative Program, Mar. 2006, Anaheim, CA. [Online]. Available: www.osa.org/partner/ network/program/Ryan%20John.ppt [15] B. Fabianek, “Optical networking testbeds in Europe,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper OWU1. [16] K. Kitayama, T. Miki, T. Morioka, H. Tsushima, M. Koga, K. Mori, S. Araki, K. Sato, H. Onaka, S. Namiki, and T. Aoyama, “Photonic network R&D activities in Japan—Current activities and future prospects,” J. Lightw. Technol., vol. 23, no. 10, pp. 3404–3418, Oct. 2005. [17] D. T. Neilson, D. Stiliadis, and P. Bernasconi, “Ultra-high capacity optical IP routers for the networks of tomorrow: IRIS project,” presented at the Eur. Conf. Optical Commun. (ECOC), Glasgow, U.K., Sep. 2005, Paper We 1.1.4. [18] D. J Blumenthal, “LASOR (label switched optical router): Architecture and underlying integration technologies,” in Proc. ECOC, Glasgow, U.K., Sep. 2005, p. 49. [19] [Online]. Available: http://www.nlr.net/ess/index.html [20] P. Freeman, (2006, Mar.)“GENI: Global environment for networking innovations,”Proc. Future InternetParis, France.OECD. [Online]. Available: http://www.oecd.org/dataoecd/43/63/36274169.pdf [21] J. Lin and J. Wu, “Optical networking testbeds in China,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper OWU2. [22] L. Zong, P. Ji, T. Wang, O. Matsuda, and M. Cvijetic, “Study on wavelength cross-connect realised with wavelength selective switches,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper NThC3. [23] E. Le Rouzic and S. Gosselin, “160 Gb/s optical networking: A prospective techno-economic analysis,” J. Lightw. Technol., vol. 23, no. 10, pp. 3024–3033, Oct. 2005. [24] P. Torab, V. Hutcheon, and D. Walters, “Waveband switching efficiency in WDM networks: Analysis and case study,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 7, 2006, OTuG. [Optical Network Architectures]. [25] E. Hernandez-Valencia and H. Menendez, “Exploiting carrier Ethernet to deliver profitable new services,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper NTUD1. [26] D. Simeonidou, R. Nejabati, B. St. Arnaud, M. Beck, P. Clarke, B. Hoang, D. Hutchison, G. Karmous-Edwards, T. Lavian, J. Leigh, J. Mambretti, V. Sander, J. Strand, and F. Travostino, Optical Network Infrastructure for Grid . [Online]. Available: http://www.ggf.org/gf/docs/?final [27] D. Payne and R. Davey, “Long reach access networks—Avoiding the bandwidth–price dilemma,” presented at the Broadband Europe Conf., Bruges, Belgium, Dec. 2004, Paper T01A.02. GGF. [28] K. Kitayama, X. Wang, and N. Wada, “A solution path to gigabitsymmetric FTTH: OCDMA over WDM PON,” presented at the Broadband Europe Conf., Bordeaux, France, Dec. 2005. paper T01A.02. [Online]. Available: http://www.bbeurope.org [29] R. C. Mendez, P. Toliver, S. Galli, A. Agarwal, T. Banwell, J. Jackel, J. Young, and S. Etemad, “Network applications of cascaded code translation for WDM-compatible spectrally phase encoded optical CDMA,” J. Lightw. Technol., vol. 23, no. 10, pp. 3129–3231, Oct. 2006. [30] M. Brown, “Blueprint for the future of high performance networking,” Commun. ACM (CACM), vol. 46, no. 11, pp. 30–77, Nov. 2003. ASON. [31] [Online]. Available: http://www.gridforum.org/ [32] [Online]. Available: http://www.uclp.ca/ [33] A. Sahara, R. Kasahara, E. Yamazaki, S. Aisawa, and M. Koga, “The demonstration of congestion controlled optical burst switching network

4695

[34] [35] [36]

[37]

[38]

[39]

[40] [41]

[42] [43]

[44]

[45] [46] [47]

[48] [49]

[50] [51]

[52]

[53] [54]

[55] [56]

utilizing two-way signaling field trial in JGN II testbed,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2005, Paper OFA7. H. Guo, Z. Lan, J. Wu, and J. Lin, “A testbed for optical burst switching network,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2005, Paper OFA6. [Online]. Available: http://www.matissenetworks.com/product.php D. Simeonidou, R. Nejabati, G. Zervas, D. Klonidis, A. Tzanakaki, and M. J. O’Mahony, “Dynamic optical network architectures and technologies for existing and emerging grid services,” J. Lightw. Technol., vol. 23, no. 10, pp. 3347–3357, Oct. 2005. F. Xue, Z. Pan, H. Yang, J. Yang, J. Cao, K. Okamoto, S. Kamei, V. Akella, and S. J. B. Yoo, “Design and experimental demonstration of a variablelength optical packet routing system with unified contention resolution,” J. Lightw. Technol., vol. 22, no. 11, pp. 2570–2581, Nov. 2004. H. N. Poulsen, D. Wolfson, S. Rangarajan, and D. J. Blumenthal, “Burst mode 10 G optical header recovery and lookup processing for asynchronous variable length 40 G optical packet switching,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper OFJ2. X. Wang and N. Wada, “Demonstration of OCDMA traffic over optical packet switching networks with PLC and SSFBG en/decoders for time domain OC processing,” presented at the Eur. Conf. Optical Commun. (ECOC), Glasgow, U.K., Sep. 2005, PDL We1.4.5. K. Stubkjaer, “Semiconductor optical amplifier-based all-optical gates for high speed optical processing,” IEEE J. Sel. Topics Quantum Electron., vol. 6, no. 6, p. 1428, Nov. 2000. G. Maxwell, R. McDougall, R. Harmon, M. Neild, L. Rivers, and A. Poustie, “WDM-enabled, 40 GB/s hybrid integrated all-optical regenerator,” presented at the Eur. Conf. Optical Commun. (ECOC), Glasgow, U.K., Sep. 2005, PDL 4.2.2. S. Nakamura, Y. Ueno, and K. Tajima, “168 Gb/s all-optical wavelength conversion with a symmetric-Mach–Zehnder-type switch,” IEEE Photon. Technol. Lett., vol. 13, no. 10, p. 1091, Oct. 2001. S. Pau, Y. Jianjun, K. Kojima, N. Chand, and V. Swaminathan, “160-Gb/s all-optical MEMS time-slot switch for OTDM and WDM applications,” IEEE Photon. Technol. Lett., vol. 14, no. 10, pp. 1460–1462, Oct. 2002. D. Petrantonakis, G. T. Kanellos, P. Zakynthinos, N. Pleros, D. Apostolopoulos, and H. Avramopoulos, “A 40 Gb/s 3R burst-mode receiver with 4 integrated MZI switches,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper PDP 25. P. Koonath, T. Indukuri, and B. Jalali, “Monolithic 3-D Silicon Photonics,” J. Lightw. Technol., vol. 24, no. 4, pp. 1796–1804, Apr. 2006. [Online]. Available: http://www.glimmerglass.com R. Varrazza, I. B. Djordjevic, and Y. Siyuan, “Active vertical-couplerbased optical crosspoint switch matrix for optical packet-switching applications,” J. Lightw. Technol., vol. 22, no. 9, pp. 2034–2042, Sep. 2004. [Online]. Available: http://www.yokogawa.com V. Van, A. Ibrahin, P. P. Absil, F. G. Johnson, R. Grover, and P. Ho, “Optical signal processing using nonlinear semiconductor microring resonators,” IEEE J. Sel. Topics Quantum Electron., vol. 8, no. 3, pp. 705–713, May 2002. C. Gobby, Z. L. Yuan, and A. J. Shields, “Quantum key distribution over 122 km of standard telecom fiber,” Appl. Phys. Lett., vol. 84, no. 19, pp. 3762–3764, May 2004. A. Argyris, D. Syvridis, L. Larger, V. Annovazzi-Lodi, P. Colet, I. Fischer, J. García-Ojalvo, C. R. Mirasso, L. Pesquera, and K. A. Shore, “Chaos-based communications at high bit rates using commercial fibreoptic links,” Nature, vol. 438, no. 7066, pp. 343–346, Nov. 17, 2005. Y. Liu, E. Tangdiongga, Z. Li, H. de Waardt, A. M. J. Koonen, G. D. Khoe, H. J. S. Dorren, X. Shu, and I. Bennion, “Error-free 320 Gb/s SOA-based wavelength conversion using optical filtering,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper PDP 28. O. Lecrec et al., “Optical regeneration at 40 Gb/s and beyond,” J. Lightw. Technol., vol. 21, no. 11, pp. 2779–2790, Nov. 2003. T. Akiyama et al., “Application of spectral-hole burning in the inhomogeneously broadened gain of self-assembled quantum dots to a multichannel nonlinear optical device,” IEEE Photon. Technol. Lett., vol. 12, no. 10, pp. 1301–1303, Oct. 2000. N. Beheshti, Y. Ganjali, R. Rajaduray, D. Blumenthal, and N. McKeown, “Buffer sizing in all-optical packet switches,” presented at the Optical Fiber Commun. Conf. (OFC), Anaheim, CA, Mar. 2006, Paper OThF8. C. Chang-Hasnain, “Controlling light speed in semiconductors,” presented at the Eur. Conf. Optical Commun. (ECOC), Glasgow, U.K., Sep. 2005, Paper Mo3.1.3.

4696


Michael J. O’Mahony (M’91–SM’91) received the Ph.D. degree in digital transmission systems from the University of Essex, Colchester, U.K., in 1978. In 1979, he was with the Optical System Research Division, British Telecom, working on research into fiber-optic systems for undersea systems, in particular, experimental and theoretical studies of receiver and transmitter design. In 1991, he joined the Department of Electronic Systems Engineering, University of Essex, as Professor of communication networks and was Head of the department from 1996 to 1999. He is the Principal Investigator for grants supported by industry, national research councils, and the EU. He is the author of over 200 papers relating to optical communications. His current research is related to the study of future network infrastructures and technologies. Dr. O’Mahony is a member of the Institution of Electrical Engineers (IEE).

Christina Politi (S’05) was born in Zografou, Greece, in 1975. She received the degree in physics from the University of Athens, Athens, Greece, in 1998 and the M.Sc. degree in the physics of laser communications from the University of Essex, Colchester, U.K., in 2000. She is currently working toward the Ph.D. degree on ultrafast wavelength conversion and optical processing for optical circuit and packet-switched networks at the University of Essex. She has been involved with the European IST-OPTIMIST and IST-BREAD projects.

Dimitrios Klonidis (M’02) received the degree in electrical and computer engineering from Aristotle University, Thessaloniki, Greece, in 1998 and the M.Sc. degree in telecommunication and information systems from the University of Essex, Colchester, U.K., in 2001. He is currently working toward the Ph.D. degree in high-speed optical packet switching at the University of Essex. Since August 2004, he has worked for the EPSRC project OPSnet. He is currently involved with a number of research activities within the University of Essex. His main research topics are in the area of ultrafast optical communication networks.

Reza Nejabati (M’02) received the B.Sc. degree in electrical engineering from Shahid Beheshti (Melli) University, Tehran, Iran, in 1997 and the M.Sc. degree (with distinction) in telecommunication and information systems from the University of Essex, Colchester, U.K., in 2001. He is a Senior Research Officer with the Photonic Network Research Group, University of Essex. His main research interests are in the area of optical subwavelength switching and grid networking.

Dimitra Simeonidou (M’95) received the B.Sc. and M.Sc. degrees from the Physics Department, Aristotle University, Thessaloniki, Greece, in 1987 and 1989, respectively, and the Ph.D. degree from the University of Essex, Colchester, U.K., in 1994. She has over ten years of experience in the field of optical transmission and optical networks. From 1992 to 1994, she was a Senior Research Officer with the University of Essex in association with the MWTN RACE project. In 1994, she was a Principle Engineer with Alcatel Submarine Networks and contributed to the introduction of wavelength-division-multiplexing (WDM) technologies in submerged photonic networks. She participated in standardization committees and was an advising member of the Alcatel Submarine networks patent committee. She is a currently a Professor with the University of Essex. She is the author over 120 papers and holds 11 patents relating to photonic technologies and networks. Her main research interests include optical wavelength and packet-switched networks, network control and management, and GRID networking.