Wavelength Division Multiplexing

Transcription

1 Wavelength Division Multiplexing Yassine Khlifi and Noureddine Boudriga, Carthage University, Tunisia Mohammad S. Obaidat, Monmouth University, New Jersey Introduction 1 Optical Multiplexing Techniques 2 Major Optical Components 2 Optical Multiplexing 4 Synchronous Optical Network Technique 5 Synchronous Digital Hierarchy 6 Wavelength Division Multiplexing Technology 6 Basic Concepts 6 WDM Evolution, Advantages, and Limits 7 WDM Survivability 7 Dense WDM 8 WDM Switching Techniques 9 Wavelength Routing 9 Optical Packet-Switching Technique 10 Optical Burst-Switching Technique 11 Optical Label-Switching Technique 12 WDM Signaling Protocols 13 OBS Signaling Schemes 13 OLS Signaling Schemes 14 WDM Network Architectures 15 Broadcast and Select Networks 15 Wavelength Routing Networks 16 QOS Requirements in WDM NETWORKS 16 Basic Concepts 16 Contention Resolution in OPS and OLS Networks 17 Contention Resolution in OBS Networks 17 QoS Parameters 18 QoS Models 18 QoS Mechanisms 19 Advanced Issues 20 Conclusion 20 Glossary 20 Cross-References 21 References 21 INTRODUCTION In recent years, the increasing popularity of the Internet has resulted in a rapidly growing demand for network bandwidth. It is expected that traditional networks will not be able to support the massive traffic resulting from emerging high-bandwidth applications. The demand of multimedia applications is creating an incredible pressure on existing telephony infrastructures that are designed to handle predictable connection-oriented voice traffic. Users are already seeing reduced performance of the Internet, including high delays and network failures. Hence, new technologies were needed to support the bandwidth requirements of subscribers and provide support for new services. Optical fiber technology is becoming the dominant factor for meeting these demands because of its potential capabilities: huge bandwidth, low signal attenuation, low power requirement, low material usage, small space requirement, and low cost. With optical technology, these networks are becoming the appropriate infrastructure for Internet transport. Given the large difference between optical transmission capacity (in tens of Tbps) and electronic data rate (several Gbps), it is expected that optical fibers can scale the increasing volume of information (Papadimitriou et al ). Currently, links in long-life networks consist almost entirely of optic fibers, with hundreds of thousands of miles of new fiber being deployed every year. Optic fiber offers increased capacity by allowing higher transmission rates over longer distances as compared to copper wires. However, the bandwidth is still limited by the electronic processing and switching speeds at the communication nodes. Three technologies are available in an optical network: time division multiplexing (TDM), synchronous optical network (SONET), and synchronous digital hierarchy (SDH). Currently, the SONET-SDH layer has been used with many traditional networks to offer mechanisms for efficient bandwidth utilization and to provide protection. The SONET-SDH framing structure is widely used in voice transmission in optical networks but is considered to be cumbersome for high-speed links. A new flexible framing scheme is required for next generation optical network to support multiprotocol encapsulation, virtual links, quality of service (QoS) differentiation, and link management (Papadimitriou et al ; SONET overview, undated). The fiber bandwidth can be further exploited by dividing the bandwidth into a number of channels of different wavelengths and by operating each channel at peak electronic rate. This technique is known as wavelength division multiplexing (WDM). A WDM network not only provides a huge amount of bandwidth but also may offer data transparency in which the network may carry signals of arbitrary format. Data transparency may be achieved through all-optical transmission and switching of signals without electronic conversions at the intermediate nodes. The major challenge in deploying WDM is to develop architectures and protocols that take full advantage of the benefits offered by WDM. At the same time, the proposed architectures must consider the limitations imposed by current optical device technology. Also, these architectures must support critical services such as bursty traffic while efficiently utilizing the network resources. There are several optical technologies, namely, wavelength routing (WR), optical packet switching (OPS), optical label switching (OLS) and optical burst switching (OBS). In WR, end users communicate with one another via all-optical WDM 1 bid44582_ch40.indd 1 7/16/07 9:07:02 PM

2 2 WAVELENGTH DIVISION MULTIPLEXING channels, which are referred to as lightpaths. An OPS is capable of dynamically allocating network resources with fine packet-level granularity while offering high scalability. An OLS network is proposed to provide a fast switching process based on a set of labels built during the signaling step. An OBS is designed to achieve a balance between the coarse-grained WR and the fine-grained OPS. In this chapter, we investigate the optical components, optical multiplexing techniques, network architectures, and protocols for WDM networks with the goal of addressing the different issues related to the existing approaches and developed architectures that provide improved performance as well as support for QoS. This chapter is organized into eight sections. The next section discusses optical multiplexing techniques and major trends in optical components. The third section presents the basic concepts of the WDM technology, as well as its advantages and limitations. The fourth section addresses the different techniques used by optical switching in the WDM networks. The fifth section presents several signaling schemes used by WDM. The sixth section provides a survey of the basic WDM architectures such as broadcast and select and WR networks. The seventh section discusses the QoS models adopted in WDM networks. Finally, the last section before the conclusion gives several critical issues related to the provision of the Internet protocol over WDM. OPTICAL MULTIPLEXING TECHNIQUES Bandwidth requirements are expected to increase with time. Therefore, research in optical fiber networks is directed toward achieving high-capacity optical networks. High capacities in optical networks can be achieved by using various optical components and multiplexing techniques. In all-optical networks, concurrency can be provided through time slots (TDM or optical TDM), SONET-SDH, and WDM. In this section, we present an overview of the major optical multiplexing techniques. Major Optical Components To transmit data across an optical fiber, the information must first be encoded, or modulated, onto the laser signal (the optical carrier). Analog techniques include amplitude modulation, frequency modulation, and phase modulation. Digital techniques include amplitude shift keying, frequency shift keying, and phase shift keying (Brackett 1990 ). To enable the operation of the new, vast, and versatile wavelength dimension, new technologies have to be developed. New components that can exercise selection, switching, and routing based on wavelengths were also needed, including optical transmitters, optical receivers, optical amplifiers, optical cross-connects (OXC) and optical add-drop multiplexers (OADMs). Optical Fiber Before describing the optical components, it is essential to understand the characteristics of the optical fiber itself. An optical fiber consists of a central core that is surrounded by a cladding layer whose refractive index is lower than the index of the core. An optical fiber acts as Figure 1: The low-attenuation regions of an optical fiber a dielectric waveguide to confine light within its surfaces and to guide it in a direction parallel to its axis. The confinement of light is achieved using the so-called total internal reflection, which is done when the angle of incidence of light is greater than the critical angle. Fiber possesses many characteristics that make it an excellent medium for high-speed networking. Figure 1 shows the two lowattenuation regions of optical fiber. The figure also shows the loss attenuation for an optical network in two regions centered at approximately 1300 nanometers (nm) and 1550 nm, respectively, where the attenuation is less than 0.5 db per kilometer. The total bandwidth for each region is approximately 25 THz. By using these attenuation areas for data transmission, the signal loss can be made extremely small, thus reducing the number of needed amplifiers and repeaters. Besides its enormous bandwidth and low attenuation, fiber also offers low error rates. Fiber-optic systems typically operate at bit error rates (BERs) of less than 10 to 11. Also, fiber transmission is immune to electron magnetic interference and does not cause interference. Naturally, the fiber also has some weaknesses, such as dispersion, nonlinear refraction, and attenuation (when signal attenuation is higher than 0.2 db/km). The major forms of dispersion are modal, polarizationmode, and chromatic. Modal dispersion occurs only in multimode fibers in which a signal propagates at different speeds. A single mode fiber carries two polarization modes that are indistinguishable in an ideal fiber because of the cylindrical symmetry of the core. However, real fiber deviates from cylindrical symmetry to a more elliptical shape because of the manufacturing process, mechanical stress applied to the fiber, or both. This accidental loss of symmetry results in two distinct polarization modes with different propagation modes. Polarization mode dispersion occurs because two polarization modes arrive at different times at the receiver. Chromatic dispersion is the consequence of two contributing factors: material dispersion and waveguide dispersion. The first results from the refractive index of silica, which is a function of the spectral components of the signal. The second factor occurs bid44582_ch40.indd 2 7/16/07 9:07:08 PM

3 OPTICAL MULTIPLEXING TECHNIQUES 3 because part of the signal s power propagates on the core and part in the cladding, and they have different refractive indexes. Optical Transmitter Optical transmitters can be either tunable or fixed. Tunable lasers can be characterized by the tuning range, tuning time, and information about whether the laser is continuously tunable or discretely tunable. The tuning range is the range of wavelengths over which the laser may be operated. The tuning time is the time required for the laser to tune from one wavelength to another. The most popular tunable lasers are mechanically tuned lasers: acousto-optically and electro-optically tuned lasers or injection current tuned lasers. Mechanically tuned lasers use a Fabry-Perot cavity that is adjacent to the lasing medium to filter out unwanted wavelengths. They can be tuned by physically adjusting the distance between two mirrors on either end of the cavity so that only the desired wavelength interferes with its multiple reflections in the cavity. A major drawback of mechanically tuned lasers is the tuning time. Because of the mechanical nature of tuning and the length of the cavity, the tuning time is on the order of milliseconds. In acousto-optically and electro-optically tuned lasers, the index of refraction in the external cavity is changed by using sound waves and electrical current, respectively. The change in the index of refraction results in the transmission of light at different frequencies. In tuned lasers, the tuning time is limited by the time required for light to get back in the cavity at the new frequency (Brackett 1990 ). Injection current tuned lasers allow wavelength selection via diffraction grating. Normally, the grating consists of a waveguide in which the index of refraction alternates periodically between two values. The wavelengths that match the period and indexes of the grating are constructively reinforced and propagated through the waveguide, whereas the other wavelengths are not propagated. Depending on the placement of the grating, a laser can be either distributed feedback (DFB) or distributed Bragg reflector (DBR) laser. In DFB, the grating is placed in the lasing medium, whereas in DBR it is placed outside of the lasing medium. The tuning time for the DBR laser is less than 10 ns. Optical Receiver An optical receiver can be either tunable or fixed. The most popular tunable filters are Etalon, acousto-optic, electrooptic, and liquid crystal (LC) Fabry-Perot ( Intellilight dedicated SONET ring, undated). The Etalon comprises a single cavity formed by two parallel mirrors. A signal from an input fiber enters the cavity and reflects a number of times between the two mirrors. A single wavelength can be chosen for propagation through the cavity by adjusting the distance between the mirrors, which can be done mechanically or by changing the index of the material within the cavity. An example of a mechanically tuned Etalon is the Fabry-Perot Etalon receiver (Brackett 1990 ). In acousto-optic filters, radio frequency waves are passed through a transducer, which is a piezoelectric crystal that converts sound waves to mechanical movement. The sound waves change the crystal s index of refraction and enable the crystal to act as a grating. By changing the radio waves, a single optical wavelength can be chosen to pass through the filter. The tuning range for acoustooptic receivers covers the spectrum from 1300 nm to 1560 nm and allows approximately one hundred channels. The tuning time of the filters is approximately 10 ms (Yao et al ; Papadimitriou et al ). Electrodes located in the crystal are used to supply current to the crystal. The current changes the crystal s index of refraction, which in turn allows some wavelengths to pass. The tuning time and tuning range of these filters are 10 ns and 16 nm, respectively. In LC Fabry-Perot filters, the cavity consists of a liquid crystal. Electrical current is used to change the refractive index of the liquid crystal to allow some wavelengths to pass, just as in electro-optic filters. The tuning time of these filters is on the order of a few microseconds while tuning ranges in the order of 30 nm to 40 nm. Fixed receivers use fixed filters or grating devices to filter out wavelengths from a set of wavelengths in a single fiber. The most popular filters or grating devices are diffraction grating, fiber Bragg grating, and thin-film interference filters. The diffraction grating is typically a flat layer of transparent material such as glass or plastic with a row of parallel grooves cut into it. The grating separates light into its component wavelengths by reflecting it with the grooves at all angles. At certain angles, only one wavelength adds constructively while all other wavelengths interfere destructively. Thus, a desired wavelength can be selected by tuning the filter to that wavelength. In fiber Bragg grating, a periodically variable index of refraction is directly photoinduced into the core of an optical fiber. The Bragg grating reflects a given wavelength of light back to the source and passes the other wavelengths. There are two major drawbacks of this method: (1) It induces a grating directly into the core of a fiber, which leads to low insertion loss; and (2) the refractive index in the grating varies with temperature (e.g., an increase in temperature reflects longer wavelengths) (Simmons, Goldstein, and Saleh 1999 ; Yao et al ; Papadimitriou et al ). Thin-film interference filters are similar to fiber-grating devices, but they are made by placing alternate layers of low-index and high-index materials onto a substrate layer. Major disadvantages of these filters are their poor thermal stability, high-insertion loss, and poor spectral profile. Optical Amplifier Optical amplifiers are important components in the fiber links. They are used to compensate for the attenuation in fibers and for insertion loss in passive optical components such as OXC and OADM. The amplification is usually achieved using erbium-doped fiber amplifiers (EDFAs). An EDFA basically consists of small-length optical fibers doped with erbium; they range in length from a few meters to 10 meters. The erbium atoms in the fiber are pumped from their ground state to an excited state at a higherenergy level using a pump source. An incoming signal photon triggers these atoms to come down to their ground state. Thus, incoming signal photons trigger the additional photons, resulting in optical amplification (Brackett 1990 ). bid44582_ch40.indd 3 7/16/07 9:07:10 PM

4 4 WAVELENGTH DIVISION MULTIPLEXING An optical amplifier is used to amplify a weak or distorted signal with the aim of generating a better optical signal quality. It operates in the optical domain without converting the signal into electrical pulses. It is usually used in long-haul networks where the cumulative loss is large. Current amplifier systems provide extremely low noise and flatter gain, which is advantageous to the optical system. The amplifier output power has steadily increased to nearly 20 db, which is many times more powerful than the primitive models. A major advantage of EDFAs is that they are capable of amplifying signals with many wavelengths simultaneously. This has provided another way of increasing the capacity: Rather than increasing the bit rate, the bit rate is kept the same and more than one wavelength is used (Yao et al ). Optical Cross-Connect A fiber-optical cross-connect element switches optical signals from input ports to output ports. The basic OXC element is the 2 2 cross-point element. It routes optical signals from two input ports to two output ports and has two states: cross state and bar state. In the cross state, the signal from the upper input port is routed to the lower output port, and the signal from the lower input port is routed to the upper output port. In the bar state, the signal from the upper input port is routed to the upper output port, and the signal from the lower input port is routed to the lower output port. Optical cross-point elements have been used in two technologies: (1) the generic directive switch, in which light is physically directed to one of two different outputs; and (2) the gate switch, in which optical amplifier gates are used to select and filter input signals to specific output ports (National Communications System 2002 ). Directive switches can be viewed as directional couplers with no gap between the waveguides in the interaction region. When properly fabricated, both cross and bar states can be electro-optically achieved with good crosstalk performance. Cross talk is caused by interference of signals between different wavelengths (interband cross talk) or by interference on the same wavelength on another fiber (intraband cross talk). Interband cross talk must be considered when determining channel spacing where intraband cross talk occurs in switching nodes in which multiple signals on the same wavelength are being switched from different inputs to different output ports. Other types of switches include the mechanical-optical fiber switch and the thermo-optic switch. These devices offer slow switching (in milliseconds) and may be employed in circuit-switched networks. Thermo-optic waveguide switches are operated by the use of the thermo-optic effect. Gate switches operate around 1300 nm and have an optical bandwidth of 40 nm, low polarization dependence (1 db), and fairly low cross talk (below 40 db). Optical Add-Drop Multiplexers Optical add-drop multiplexers (OADM) are devices that are used to add or remove single wavelengths from a fiber without disturbing the transmission of other signals (see Figure 2 ). OADM is required for WDM ring and bus networks in order to link the network with local transmitters and receivers. OADMs provide interconnection between network structures. They are generally evaluated in terms Demultiplexer λ 1 λ i λ w of performance through cross-talk measurements. Optical cross talk at the same wavelength as the transmitted signal is generally referred to as homodyne or in-band cross talk. It is particularly serious because it cannot be removed by filtering and has been shown to severely limit network performance (Simmons, Goldstein, and Saleh 1999 ). Within homodyne cross talk, incoherent cross talk causes rapid power fluctuations, while coherent cross talk changes the optical power of the signal. Incoherent cross talk occurs in ring and bus networks when the signal and interferer are from different optical sources. It causes power variations at the receiver, resulting in BER degradation. Coherent cross talk in these networks results from multiple paths between ports within OADM. It causes variable attenuation levels between OADM ports. There is no accompanying power penalty because the BER is measured against the optical power at the receiver. Incoherent and coherent cross talk together give a range of possible power penalties, because coherent cross talk can cause variation in both signal and incoherent cross-talk powers at the receiver. The combination of coherent and incoherent cross talk leads to a range of possible BER and power penalties for OADM deployed in a network link. Optical Multiplexers and Demultiplexers Optical multiplexers and demultiplexers are used to combine multiple wavelengths onto a single fiber and allow all of the signals to be transferred using the same fiber. These optical components can increase the offered fiber capacity without adding more fibers. They also can receive an optical composite signal that consists of multiple optical frequencies from a fiber and separate it into its frequency components, which are directed to separate fibers. In addition, they can serve as an access point to the optical layer in many more aspects, including other optical devices such OADM and OXC. Optical Multiplexing Multiplexer Figure 2: Optical add-drop multiplexer The need for multiplexing is driven by the fact that it is much more economical to transmit data at higher rates over a single fiber than it is to transmit at lower rates over multiple fibers. Multiplexing and demultiplexing aim to transmit multiple signals over a single communication bid44582_ch40.indd 4 7/16/07 9:07:10 PM

5 OPTICAL MULTIPLEXING TECHNIQUES 5 channel. The two common multiplexing techniques are FDM, which separates signals by modulating the data onto different carrier frequencies, and TDM, which separates signals by interleaving bits, one after another. There are three possible multiplexing techniques: (1) time division multiplexing, (2) SONET-SDH and (3) wavelength division multiplexing. TDM is a scheme that combines numerous signals for transmission on a single communications line or channel. Each communication channel is divided into many time segments, each having extremely short duration. A multiplexer at the source of a communication link accepts the input from each individual end user, divides each signal into segments, and assigns the segments to time slots in a rotating sequence. Optical TDM (OTDM) is a promising multiplexing technique. It is similar to the electronic TDM. The only difference is that OTDM is faster and the devices used have an optical nature ( IP-over-SONET configuration 2002 ). The operational principle of OTDM is to interleave several signals in time and then direct them to the same fiber. The main advantages of multiplexing are the facts that only one source is required and that the node equipment is simpler in the single channel architecture. In OTDM, many lower-speed data channels each transmit in the form of ultrashort-duration optical pulses that are timeinterleaved to form a single high-speed stream. The resulting data stream is then transmitted over an optical fiber. Special considerations are required to generate ultrashortduration optical pulses. In particular, gain-switched semiconductor lasers and mode-locked lasers are currently used to generate such optical pulses. However, a gain-switched semiconductor laser has limitations such as the spectral spread, which results from high chirp rates and nonnegligible levels of timing jitter. These limitations can be overcome by using optical fibers with appropriate dispersion compensation and optical filtering. Synchronous Optical Network Technique SONET is a standard for optical telecommunications transport as defined by the American National Standards Institute. The standard has also been integrated into the synchronous digital hierarchy recommendations of the International Telecommunications Union (ITU) (formerly Consultative Committee on International Telegraph and Telephone, or CCITT) ( IP-over-SONET configuration 2002 ). In SONET, the basic signal rate is Mbps. This rate is known as the synchronous transport signal level 1 (STS-1) rate. Similar to other physicallayer transport, SONET describes transmission speed, line encoding, and signal multiplexing. SONET defines this as optical carrier (OC) signals, frame format, and operations, administration, maintenance, and provisioning (OAM&P) protocol. The fiber-optic transmission rates from OC-1 through OC-192. They are shown in Table 1. SONET divides a fiber path into multiple logical channels called tributaries. A tributary s basic unit of transmission is a STS-1 or OC-1 (optical carrier level 1) signal. Both operate at Mbps; STS describes electrical signals, and OC refers to the same traffic when it is converted into optical signals. SONET also allows channels to be multiplexed. Therefore, an OC-12 circuit, for instance, carries traffic from four OC-3 links. A circuit also can carry a single channel. To further understand the elements of SONET, key concepts need to be discussed. Frame format structure is based on the STS-1 equivalent to Mbps. Higher levels of signals are integer multiples of the base rate, Mbps (Papadimitriou et al ; IP-over- SONET configuration 2002 ; SONET graphical overview, undated). The STS-1 signal is divided into two main areas: transport overhead and synchronous payload overhead (SPE). The SPE is further divided into two overheads: the STS path overhead (POH) and the payload. The payload is the data being transported and switched through the SONET network without being demultiplexed at the terminating locations. STS-1 has a sequence of 810 bytes (6480 bits), which includes overhead bytes and the envelope capacity for transporting payload. The frame consists of a 90 column by 9 row structure with a frame length of 125 ms. This means that 8000 frames are sent per second, because 9 90 bytes/frame 8 bits/byte 8000 frames/s Mbps. The order of transmission of bytes is row-by-row from top to bottom and from left to right. Table 1: Fiber Optical Transmission Rates OC level Bit rate (Mbps) S0 Number DS-1 Number DS-3 Number bid44582_ch40.indd 5 7/16/07 9:07:12 PM

6 6 WAVELENGTH DIVISION MULTIPLEXING The STS payload pointer provider contained in the transport overhead designates the location of the byte where the STS-1 SPE begins. The STS POH is used to communicate various data such as the pickup and dropoff points. The higher STS- N signals are accomplished through byte interleaving of STS-1 modules. SONET provides overhead information that allows simpler multiplexing and greatly expanded operations, administration, maintenance. and provisioning capabilities. Synchronous Digital Hierarchy The deployment of synchronous transmission systems is characterized by their ability to interwork with existing plesiochronous systems. Advances in these systems have lead to plesiochronous digital hierarchy (PDH), which is a multiplexing technique that allows for combining slightly nonsynchronous rates. PDH has evolved in response to the demand for basic voice telephony and, as such, is not suited for the efficient delivery and management of highbandwidth connections. For this reason, synchronous transmission has been introduced to overcome these limits ( Synchronous optical network, undated). Synchronous digital hierarchy defines a structure that enables plesiochronous signals to be combined together and encapsulated within a standard SDH signal. The CCITT recommendations defined a number of basic transmission rates within SDH. The first of these is 155 Mbps, normally referred to as STM-1 (where STM stands for synchronous transport module). Higher transmission rates of STM-4, STM-16 and STM-64 (622 Mbps, 2.4 Gbps, and 10 Gbps, respectively) are also defined. The recommendations also define a multiplexing structure whereby an STM-1 signal can carry a number of lower rate signals as payload, thus allowing existing PDH signals to be carried over a synchronous network ( SONET graphical overview, undated). SDH defines a number of containers, each corresponding to an existing plesiochronous rate. Information from a plesiochronous signal is mapped into the relevant container. Each container then has some control information known as the path overhead added to it. The path overhead bytes allow the network operator to achieve end-path monitoring such as error rates. Together, the container and the path overhead form a virtual container (VC), which has its own frame structure made of nine rows and 261 columns. The first column is called path overhead. The payload container, which can itself carry other containers, follows it. Virtual containers can have any alignment within the administrative unit; the pointer in row 4 indicates this alignment. Within the section overhead, the first three rows are used for the regenerator section overhead, and the last five rows are used for the multiplex section overhead. Earlier SDH supports a concept called virtual containers ( SONET graphical overview, undated; Synchronous optical network, undated). Through the use of pointers, VCs can be carried in the SDH payload as independent data packages. In a synchronous network, all equipment is synchronized to an overall network clock. Notice, however, that the delay associated with a transmission link may vary slightly with time. As a result, the location of virtual containers within an STM-1 frame may not be fixed. Associating a pointer with each VC accommodates these variations. The pointer indicates the position of the beginning of the VC in relation to the STM-1 frame. It can be incremented or decremented if needed to accommodate the position of the VC. There are different combinations of virtual containers that can be used to fill up the payload area of an STM-1 frame. The process of loading containers and attaching overhead is repeated at several levels in the SDH, resulting in the aggregation of smaller VCs within larger ones. This process is repeated until the largest size of VC is filled. When filled, the large VC is loaded into the payload of a STM-1 frame. When the payload area of the STM-1 frame is full, control information bytes are added to the frame to form the section overhead. The section overhead bytes remain with the payload for the fiber section between two synchronous multiplexers. Their purpose is to provide communication channels with alignment and a number of other functions. A higher transmission rate, when required, is achieved by using a relatively straightforward byte-interleaved multiplexing scheme. Finally, let us notice that under OTDM, each end user should be able to synchronize with one time slot. The optical TDM bit rate is the aggregate rate over all TDM channels in the system, while the optical SONET-SDH rate may be higher than each user s data rate. As a result, the TDM bit rate and the SONET-SDH rate may be much higher than electronic processing speed, Thus, TDM and SONET-SDH are relatively less attractive than WDM, because WDM, unlike TDM or SONET-SDH, has no such requirement. Specifically, WDM is the current favorite multiplexing technology for long-haul communications in optical communication networks because all of the end-user equipment needs to operate only at the bit rate of a WDM channel, which can be chosen arbitrarily. WAVELENGTH DIVISION MULTIPLEXING TECHNOLOGY Wavelength division multiplexing divides the large bandwidth of a fiber into many nonoverlapping wavelength bands that can operate simultaneously with the fundamental requirement that each channel operate at a different wavelength. In this section, we introduce basic definitions of WDM and describe its technological evolution. We also highlight its advantages and limitations. We then present different protection schemes used in WDM networks. Basic Concepts WDM has been defined as a technique used in optical fiber communications, by which two or more optical signals having different wavelengths may be combined and simultaneously transmitted in the same direction over one fiber. These signals are then separated, by wavelength, at the distant end. In other words, WDM is a technology that levels optical-fiber cable in ways that allow us to have multiple available bandwidths rather than a single wavelength fiber. As illustrated in Figure 3, optical signals with wavelengths 1, 2, 3, and 4 are multiplexed and simultaneously transmitted in the same direction over a single optical fiber cable. The basic idea is based on bid44582_ch40.indd 6 7/16/07 9:07:12 PM

7 WAVELENGTH DIVISION MULTIPLEXING TECHNOLOGY 7 WDM multiplexer λ 1 λ 2 Optical amplifier WDM demultiplexer λ 1 λ 2 λ 3 λ 4 λ 1, λ 2, λ 3, λ 4 λ 3 λ 4 Figure 3: WDM transmission system WDM transmitters simultaneously transmitting several signals using different wavelengths per fiber. This way, WDM provides many virtual fibers on a single physical fiber. Today. WDM can exploit the huge optoelectronic bandwidth difference by requiring that each end user s equipment operate only at electronic rate, although multiple WDM channels from different end users may be multiplexed on the same fiber. Under WDM, the optical transmission spectrum is structured into a number of nonoverlapping wavelength bands, each supporting a single communication channel. WDM Evolution, Advantages, and Limits Recently, it has become apparent that the major part of future network traffic will be IP-based. The evolution will go toward IP-over-WDM networks, where several approaches have already been proposed (Chlamtac, Ganz, and Karmi 1992 ). Each additional layer naturally brings extra overhead to the transmission. Hence, the standard IP over ATM over SONET-SDH over WDM mapping can be considered as an inefficient solution. The other extreme is a direct IP or MPLS over WDM solution: so called -labeling or optical label (lambda) switching. It is anticipated that the next generation of the Internet will employ WDM-based optical backbones. Current development activities indicate that WDM networks will be deployed mainly as backbone networks for smaller and larger regions, as well as for metropolitan areas. End users to whom the backbone will be transparent (except for significantly improved response times) will be attached to the network through a wavelength-sensitive switching and routing node. The evolution of WDM technology has far outpaced the development of applicable standards. This has created a global concern regarding interoperability. Recently, the ITU-T has recommended that vendors use wavelengths in the spectrum range of 1520 nm to 1565 nm, although such a wide range allows for hundreds of system and product variations. Even when vendors agree to use the same wavelengths, there are still the problems of switching channels between networks. There are no immediate standards that take into account such factors as fault management, or power levels, for WDM networks (National Communications System 2002 ). WDM s main advantages include signal transparency, scalability, flexibility, and ability to upgrade fiber bandwidth. The transparency property makes it possible to support various data formats and services simultaneously. In addition to this great flexibility, transparency protects the investments with respect to future developments. WDM can support a variety of future protocols without making any changes to the network infrastructure. As networks migrate from simple point-to-point WDM to optical rings with optical add-drop multiplexers covering applications that span from metropolitan area to ultralong haul, it becomes increasingly more important to also migrate core metroregional transport infrastructures into reconfigurable, manageable, and cost-effective architectures. WDM network components such as reconfigurable OADM and OXC can be used to create WDM networks that can be operated in provisioned mode (wavelength routing) or in switched mode (packet switching, label switching, and burst switching). Because of their obvious advantages, WDM networks are rapidly deployed in long-distance carriers, local carriers, and enterprise networks. However, there are still a few limitations to be considered for WDM networks. One limitation is the nonlinearity characteristics of fiber optics. Optical fiber nonlinearities will significantly affect the performance of WDM networks. Unless corrected, these nonlinearities will lead to attenuation, distortion, and cross-channel interference. They also place constraints on the spacing between adjacent wavelength channels, reduce the maximum power of the channel, and limit the maximum bit rate. Another limiting factor is the use of wavelength converters. Wavelength conversion has been proposed for use in multihop WDM networks to improve efficiency. Wavelength converters are expensive and may not be economically justifiable because the cost is directly proportional to the number of nodes in the network. Also, several switch architectures have been proposed to allow sharing of converters among various signals at a single switch. However, experiments have shown that the performance of such a network saturates when the number of converters at a switch increases beyond a certain threshold. In addition, a certain type of converter is known to generate significant signal degradation when the output signal is converted to a higher (upconverted) signal. This produces devastating effects when a signal of a transmitted packet passes through multiple converters. However, this type appears to produce desirable results when the output signal is downconverted. WDM Survivability A fiber failure in WDM optical network causes the failure of all of the connections that traverse the failed fiber, resulting in a significant loss of bandwidth and bid44582_ch40.indd 7 7/16/07 9:07:13 PM

8 8 WAVELENGTH DIVISION MULTIPLEXING revenue. Thus, the network designer must provide a faultmanagement technique and also protect against fiberamplifier malfunction. Protection Schemes There are essentially two types of fault-management techniques to combat fiber failures in WDM networks: protection and restoration. In protection, additional capacity is reserved during connection setup, and it is used when the primary connection fails. In restoration, no extra capacity is made available during connection setup; instead, the disrupted connections are rerouted using whatever capacity is available after the fault has occurred. In this subsection, we will discuss these two fault-management strategies in WDM networks. Protection can be classified into two groups: path protection and link protection. In path protection, the traffic is rerouted through a link-disjoint backup route when a link failure occurs on its active path. In link protection, the traffic is rerouted only around the failed link. Path protection usually has fewer resource requirements and lower end-to-end propagation delays for the recovered route. In path protection, for each lightpath that is set up, there are two disjoint paths: a primary path and a backup path. The lightpath is set up on the primary path. In case of a link failure, the lightpath is switched to the prereserved backup path. The primary and the backup paths are link-disjoint, whereas the backup paths of different connections may share wavelengths on common links (Chiu and Modiano 2000 ). The switches on backup paths can be configured at the beginning that is, when the lightpath is set up on the primary path. No switch configuration is then necessary when the failure occurs. This type of recovery can be extremely fast, but the resources are not utilized efficiently. Because a protection route for each active route is preplanned, rerouting is faster and simpler than a restoration, which is usually performed in a distributed manner. Based on the availability of a dedicated protection versus a shared-protection scheme, three types of protection techniques are used in WDM: 1 1 protection, 1:1 protection, and 1: N protection. In 1 1 protection, traffic is transmitted on both paths from the source to the destination. The destination receives data from the primary path first. If there is a failure on the primary path, the destination switches over to the backup path and continues receiving. In 1:1 protection, data are normally not transmitted on the backup path. Thus, we can use the backup path to carry low-priority pre-emptable traffic. If there is a failure on the primary path, the source node is notified (by some protocol) and switches over to retransmit on the backup path. Hence, some data may be lost and the source must be able to retransmit those data. If sharing among backup paths is allowed, the switches on the backup paths cannot be configured until the failure occurs. The recovery time in this scheme is longer, but the overall resource utilization is better than the previous protection. This scheme is called 1:N protection. Restoration Schemes Restoration can be used to provide more efficient routes after the protection has been completed or additional resilience against further faults before the first fault is fixed. Usually, the restoration mechanism is slow (seconds to minutes) and can be computed on the fly by a centralized management system. There are two kind of restoration: link restoration and path restoration. With link restoration, all of the connections that traverse the failed link are rerouted around that link. The source and destination nodes of the connections traversing the failed link are unaware of the link failure. The end nodes of the failed link dynamically discover a route around the link for each wavelength in the link. On the occurrence of a failure, the end nodes of the failed link may participate in a distributed procedure to build new paths for each active wavelength on the failed link. When a new route is discovered, the end nodes of the failed link reconfigure their cross-connects to reroute that channel onto the new route. If no new routes are discovered, the relevant connection is blocked. With path restoration, when a link fails, the source node and the destination node of each connection that traverses the failed link are informed about the failure (possibly via messages from the nodes adjacent to the link failure). The source and the destination nodes of each connection independently discover a backup route on an end-to-end basis (possibly using a different wavelength channel). When a new route and wavelength channel are discovered for a connection, network elements such as wavelength crossconnects are reconfigured appropriately, and the connection switches to the new path. If no new routes (and associated wavelength) are discovered for a broken connection, that connection is blocked. Dense WDM The WDM network provides the backbone to support existing and emerging technologies with almost limitless amounts of bandwidth capacity. With today s expansion of online services (e.g., the Internet), mobile telephony, and the anticipated emergence of multimedia services, traffic demand will grow quickly, requiring a huge increase in the transport capacity of public networks. WDM and its improvement dense wave division multiplexing (DWDM) have since proven to be the most promising technology to satisfy the capacity demand in the core network by multiplexing hundreds of Gigabit channels in one fiber. The multiplexing is performed by different lasers emitting light at different wavelengths to form signals that are multiplexed onto a single fiber. A single optical fiber is capable of carrying 10 Terabits per second when DWDM technology is used (Papadimitriou et al ; Brackett 1990 ). DWDM has been proposed and deployed in many telecommunication backbones to support growing needs. DWDM systems can upgrade the channel number in optical fibers by using more power or additional signal to noise margin. Technically, WDM and DWDM are similar, but as the name implies, DWDM supports many more wavelengths. The number of wavelengths that a DWDM system can support depends on the ability of the system to accurately filter and separate them. Initial implementations of DWDM systems support either eight or sixteen wavelengths. However, current DWDM systems are capable of supporting thirty-two or sixty-four wavelengths. An end terminal with an ITU-T compliant wavelength transmitter bid44582_ch40.indd 8 7/16/07 9:07:13 PM

9 WDM SWITCHING TECHNIQUES 9 is directly connected to the DWDM system. An optical amplifier with a gain flatness of 1 db and a maximum total output power of 21.5 dbm is a key component in the DWDM system, in which the optical power of each wavelength is kept at the same level by using EDFA at the input ports of the DWDM system. The next generation of DWDM will have 128 wavelengths, a total capacity of 1.28 Tbps and even more, and networking functions such as OXC and OADM (Zang et al ). WDM SWITCHING TECHNIQUES Four major switching techniques have been proposed in the literature for transporting IP traffic over WDM-based optical networks. Accordingly, IP over WDM networks can be classified as WR networks, OPS networks, OLS networks, and OBS networks. Wavelength Routing WR networks carry data between access stations in the optical domain without any intermediate optical to and from electronic conversion. This is realized by assigning a path in the network between the two nodes and allocating a free wavelength on all links on the path. Such an alloptical path is commonly referred to as a lightpath and may span multiple fiber links without any intermediate electronic processing. Lightpath Establishment A lightpath is used to support a connection in a WR network and may span multiple fiber links. In the absence of wavelength conversion, it is required that the lightpath use the same wavelength on all fiber links of the path it uses. This requirement is referred to as the wavelength continuity constraint. Continuity may result in an inefficient utilization of WDM channels. Alternatively, when the routing nodes have limited or full conversion capability, it is possible to allocate different wavelengths on the links of the assigned path. Given a set of connections, the problem of setting up lightpaths by routing and assigning a wavelength to each connection is called the routing and wavelength assignment (RWA) problem. To establish a lightpath from source to destination, one has to determine a route along which the lightpath can be established and then assign a wavelength to the selected route. Typically, connection requests may be of two types: static and dynamic. In static lightpath establishment (SLE) problem, the entire set of connections is known in advance and the problem is then to set up lightpaths for these connections while minimizing the usage of network resources (such as the number of wavelengths or the number of fibers). In SLE algorithms, the routing procedure does not vary with time, and the routes for given source-destination pairs are predetermined based on the topology and policies criteria but independent of the current traffic condition in the network. In dynamic lightpath establishment (DLE), a lightpath is set up for each connection request as it arrives, and the lightpath is released after a small period of time. In DLE, a routing algorithm can vary with time and be adaptive in the sense that it can select a route based on the current network conditions. DLE considers alternate routing, where each source-destination pair is associated with a set of routes. If resources along one route are not available, then another route in the set is examined. When there are multiple routes to choose, the routing algorithm first tries to establish the primary path among the shortest possible routes and then tries to establish its corresponding backup paths also on the shortest possible route. The objective in dynamic traffic cases is to set up lightpaths and assign wavelengths in a manner that minimizes the amount of connection blocking or maximizes the number of connections that are established in the network. There have been extensive studies to solve both the static and the dynamic RWA problems ( Intellilight dedicated SONET ring, undated). Wavelength-routed connections are fairly static, and they may not be able to accommodate the highly variable and bursty nature of Internet traffic in an efficient manner. Traffic Grooming The important advances of high-speed transmission technology creates a large gap between the capacity of an optical channel and the bandwidth requirements of a typical connection request, which can vary in range from tens or hundreds of Mbps up to the full-wavelength capacity. Furthermore, the amount of wavelength channels available for most of the networks of practical size is lower than the number of source-destination connections that need to be made. Traffic grooming poses the problem of how to multiplex (and demultiplex) a set of low-speed traffic streams onto high-capacity channels and switch them at intermediate cross-connects (Chiu and Modiano 2000 ). The performance of WR networks depends on the efficient merging of the fractional wavelength link of the nodes into a full or almost full wavelength requirement. This merging of traffic from different source-destination pairs is called traffic grooming. Nodes that can groom traffic are capable of multiplexing and demultiplexing lowerrate traffic onto a wavelength and switching it from one lightpath to another. The grooming of traffic can be either static or dynamic. In static traffic grooming, the sourcedestination pairs, whose requirements are combined, are predetermined. In dynamic traffic grooming, connection requests from different source-destination pairs are combined, depending on the existing lightpath at the time of request arrival (Zhu and Mukherjee 2001 ). Even though WR networks have already been deployed, they may not be the most appropriate for the Internet because, for example, it takes at least a round-trip delay to establish a lightpath. This leads to poor wavelength utilization if the connection holding time is especially short. The bursty nature of the data traffic also leads to poor wavelength utilization. Therefore, to fully utilize the wavelength, a sophisticated traffic-grooming mechanism is needed to support statistical multiplexing of data from different users. The objective of traffic grooming is either to maximize the network throughput or to minimize the connection-blocking probability and improve the wavelength utilization, whereas a strategic network design problem is to minimize the total cost (Papadimitriou et al ; Chlamtac, Ganz, and Karmi 1992 ; Chiu and Modiano 2000 ). bid44582_ch40.indd 9 7/16/07 9:07:14 PM