Components of an End-to-End Optical Network

Components of an End-to-End Optical Network

At this point in the optical networking revolution, we're striving for what will be an end-to-end optical architecture. End-to-end optical architecture means that nowhere in the network is the optical signal being converted into an electronic signal. This reduction in the processing of signals would reduce costs and ultimately provide better performance.

No other transmission medium can unlock the same level of available bandwidth as can the visible light spectrum. But today, the electronic equipment acts as the bottleneck. Fibers now can carry terabits per second (Tbps), but they terminate on equipment that, at best, can handle gigabits per second (Gbps). So, before we can unleash the possibilities (and realize the savings) of end-to-end optical networking, we need to replace all the existing electronic equipment with optical equipment, which, of course, will be costly. It will involve not only new hardware but also new skill sets and network management solutions.

Components that comprise an end-to-end optical network include the following:

         Optical-line amplifiers, such as erbium-doped fiber amplifiers (EDFAs)

         Wavelength Division Multiplexing (WDM) equipment

         Optical add/drop multiplexers (OADMs)

         Optical switches

Figure 12.3 shows an example of an optical network that incorporates these components, and the following sections describe these components in detail.

Figure 12.3. Optical network components

graphics/12fig03.gif

EDFAs

As mentioned in Chapter 2, EDFAs, which were introduced in 1994, were a key innovation in the fiber world. Composed of erbium metal and doped with special atoms, EDFAs are incorporated in optical fiber at periodic intervals, generally 30 to 60 miles (50 to 100 kilometers), to boost communication signals. The components in an EDFA include an erbium-doped fiber, a laser-pump diode, couplers, and isolators. The light to be amplified is coupled in a section of erbium-doped fiber together with light from a laser-pump diode, normally about 980 nanometers. The EDFA itself operates in the range of 1,550 nanometers. The light from the laser-pump diode boosts the 1,550 nanometers light and is separated on the exit route. An isolator at either end protects the system from unwanted reflections. With EDFAs, an optical signal is amplified without having to undergo any conversions.

Before EDFAs, electronic regenerators had to extract signals, retime them, and then regenerate them. This conversion limited data rates to 2.5Gbps. EDFAs enabled us to quadruple this speed, providing data rates of 10Gbps.

WDM and DWDM

WDM works from the premise that we can spatially separate, or multiplex, different wavelengths of light down a single optical fiber. Current fiber-optic systems use only a fraction of the available bandwidth. They carry just one wavelength, when, in fact, thousands of wavelengths stand to be derived. The data rate supported by each wavelength depends on the type of light source. Today, each wavelength can carry from 2.5Gbps (that is, OC-48) to roughly 10Gbps (that is, OC-192). Some trial systems are operating at 40Gbps (that is, OC-768). In the very near future, we're expecting the delivery of Tbps light sources, and by 2005, we expect to see lasers operating in femtoseconds (that is, 10 18). With speeds in this range, the time between bursts is the same time that light takes to travel one-eighth the width of a human hair.

WDM furnishes separate channels for each service at the full rate. The idea is not to aggregate smaller channels into one larger channel, but to provide a very high-speed channel that can terminate on today's switches and routers that, in fact, support 2.5Gbps interfaces. Systems that support more than 16 wavelengths are referred to as DWDM; you'll see both acronyms WDM and DWDM used to describe this network element.

The OC-48 systems today support in the range of 60 to 160 wavelengths. OC-192 systems generally support 8 to 32 wavelengths. Researchers tell us that 320-wavelength systems are on the 2002 horizon.

On top of all this, the potential exists for transmitting thousands of channels potentially as many as 15,000 wavelengths on a single fiber with developments such as Bell Labs's "chirped-pulse WDM." The idea of chirped pulse involves a specialized mode-locked laser, which rapidly emits very wide pulses of light. Because each part of a fiber interacts differently with varying frequencies of light, the result of chirped-pulse WDM is unequal dispersion. The pulse is stretched out when it enters the fiber, and data can be put on the discrete frequencies that emerge. You can think of this process in terms of a horse race. When a race starts, horses emerge together from the gate. But because each horse keeps a separate pace, spaces soon develop between the horses. This is the same type of stretching out that happens to the laser light in chirped-pulse WMD.

If we couple the potential for 15,000 wavelengths on one fiber with each of those wavelengths supporting femtobits per second (Fbps), we have an explosion of bandwidth that is like nothing we have known before. As fantastic as all this sounds, we're likely to see even greater achievements; after all, we are still in very early stages of knowledge about what we can achieve with optical networking. Before any real progress can be made, the realm of microphotonics must develop.

DWDM Developments and Considerations

As the number of wavelengths increases and the difference between the wavelengths gets smaller, the need for wavelength stability becomes greater, to ensure that the optical carriers do not bump into each other. To get this stability, you either need network operators maintaining a stock of boards for each wavelength or you need tunable lasers.

If you have 320 wavelengths, you need 320 separate boards, each tuned to the appropriate wavelength. Obviously, for redundancy purposes, you need a backup for each of those boards. And you need this at each location where you have the WDM. So, you can see that a tunable laser that could adopt the behavior of a specific frequency, as needed, would greatly reduce the operating cost and the costs of spare parts and inventory.

Another important development is that DWDM is beginning to be able to address network survivability requirements. DWDM is also now capable of incorporating highly valued SDH/SONET-like capabilities, including monitoring performance, providing protection, and provisioning optical channels. SDH/SONET, as mentioned in Chapter 5, describes the network survivability tactic. The dual-counter-rotating rings provide a protected fiber path over which information can be shunted in the opposite direction if a fiber ring is broken. Until recently, DWDM had no such capability. It was deployed as a point-to-point link; if the fiber was cut, you lost communications between the two DWDM systems. But now we are beginning to see the introduction of those restoration capabilities onto the DWDM platforms, which means that SDH/SONET will have a more limited life in the future. Industry forecasts predict that SONET/SDH has perhaps a 10-year lifespan left, after which the benefits of DWDM will override the reliability factors that we today associate with SDH/SONET. Remember that SDH/SONET is a TDM system, and therefore it cannot take advantage of the capacity gains that DWDM systems provide.

A different consideration emerges as the DWDM systems continue to develop. Because of a combination of nonlinearities and dispersion, most of the fiber currently in place around the world possibly 95% of it would have trouble carrying the very fast (Tbps speed and Fbps pulses) signals for long distances in a DWDM system. These impairments that exist in current fiber can lead to crosstalk among the different wavelengths, interference between consecutive pulses on any signal wavelength, and degradation in the overall signal-to-noise ratio. This means that much of the fiber we have deployed over the past two decades will have to be replaced in order to take advantage of the new generation of optical equipment. Fiber solutions exist today, but it will take time and financial resources to deploy them.

Where DWDM Fits in the Network Architecture

The core network was the first place DWDM was deployed because this is where the economics made most sense. Increases in intercity traffic required carriers to expand the capacity of their long-haul pipes. So the response was to deploy these point-to-point links with DWDM. This resolved the bandwidth problem, but it did nothing to address the routing issues. WDM and DWDM currently lack the intelligence to really deliver meshed network configurations, and thus we have a need for optical switches. The main benefit of DWDM in the core is that it reduces deployment costs by eliminating the need for expensive amplifiers. Current DWDM products can operate successfully over about 300 to 450 miles (480 to 725 kilometers), and new developments are promising up to 4,000 miles (6,400 kilometers) without boosting the signal. As mentioned earlier in the chapter, the process of regenerating signals represents as much as half of the overall cost of an optical deployment. Therefore, developments in extending distances are very promising.

Another place DWDM is used is in metropolitan area networks (MANs). MANs are becoming saturated, and network expansion is costly pulling fiber along existing conduits costs about US$30,000 per mile. But traditional DWDM systems are not well suited to MANs. For one thing, they were designed to work well on point-to-point links, but MAN traffic must be dropped and added frequently. DWDM does not present the same cost justifications in the MAN as it does in the core.

The great savings that comes with DWDM in the MAN is from the reduction in the need for the expensive amplifiers. By definition, a MAN is fairly short, so there is no need for the use of expensive amplifiers. You can spend US$20,000 to US$30,000 or more for an amplifier that is capable of operating over a range of 300 to 450 miles (480 to 725 kilometers). However, runs in MANs are typically no longer than 70 miles (110 kilometers), so these expensive amplifiers are often overkill. As a result, the next generation of MAN products, designed to address the MAN core that is, metro access and the enterprise networks are being introduced. Metro core products are used for building citywide rings. Therefore, they generally support longer distances and greater capacity than do metro access products. Metro access products bring fiber closer to the customer, so they reduce deployment costs. Enterprise products address building high-capacity campus networks. In all three of these MAN sectors, the issues are the same: pricing, scalability, access, and flexibility.

As far as pricing and scalability issues go, the lower carrying capacity and distance requirements in the metro area allow providers to reduce costs by using less expensive lasers. The price of a transponder board, which represents 90% of the cost of a laser, can vary by 25%, depending on the quality of the laser. Shorter-distance lasers use less-expensive modulation and amplification techniques. Whereas long-haul lasers are externally modulated, enabling the signal to travel up to 450 miles (725 kilometers), shorter distances may allow direct modulation, where the laser runs only 50 to 60 miles (80 to 100 kilometers) but costs 30% to 40% less than a long-haul laser. But cheaper lasers also mean less capacity. Greater spacing is required between the wavelengths, thereby reducing the number of channels or wavelengths that you can derive by up to 50%.

Additional issues to be considered relative to pricing include the cost of the local loop. Eliminating the active components in the optical network can produce an even more cost-effective network. Thus, passive optical networks (PONs) are used to reduce costs by distributing costs across more endpoints and by replacing expensive OADMs or DWDM nodes with optical splitters and couplers at each fiber connection in the network. Eliminating the active components reduces the distance the signal can travel, so the theoretical range is only about 12 miles (19 kilometers). But the result can be a 10-fold saving as compared to using conventional SDH/SONET equipment, and can be even more as compared to using DWDM systems. PONs (which are covered in more detail in Chapter 13, "Broadband Access Solutions") are being considered very seriously as a means by which to deliver fiber to the home very cost-effectively.

Two major bandwidth drivers are pushing for delivery of high-speed optics down to the customer premises. First, customers are looking to connect their data centers through high-speed mainframe interfaces, such as ESCON and Fiber Channel. Second, the Internet is generating a huge demand for capacity, and it changes how traffic flows as well. As discussed in Chapter 8, "Local Area Networking," the old 80/20 scenario is reversing. It used to be that 80% of the data generated within a given business address also came from within that business address. Now, 80% of information exchange is outside the local business address. Traffic patterns also shift much more rapidly today than in the past, so they are more difficult to predict.

To meet these new demands, we need a very dynamic network that has the capability to accommodate huge capacity requirements and to change the configuration of that capacity dynamically. Subscribers want to connect at the current speed of their backbone. They want to make a direct connection through MANs and long-haul networks, with the associated protocols. And, of course, they want guaranteed QoS.

IP over DWDM

Today, bandwidth reservation and intelligent IP switches can prioritize voice and video traffic to ensure that high-priority traffic gets the first shot at the underlying bandwidth. New generations of IP-based switches provide the capability to meet QoS commitments. Layer 3/Layer 4 switching services allow the switch to prioritize and guarantee packets, based on predetermined criteria within a switch. Higher-level protocols (such as RSVP) can reserve bandwidth across an entire network. This creates a value proposition for the service provider: The ISP can deliver high bandwidth in a format that users want, for less cost, while approximating the QoS guarantees that the end user expects for high-priority traffic.

As discussed in Chapter 5, the PSTN was not built to be dynamic. It was based on a system of 64Kbps channels, or DS-0s/CEPT-0s, aggregated by time division multiplexers into DS-1/CEPT-1 or DS-3/CEPT-3 facilities that would deliver traffic into cross-connects and switches at the network core. More time division multiplexers were required at the other end to reverse the process and to distribute the DS-0s/CEPT-0s. Time division multiplexers are expensive, and they often require manual configuration, which slows provisioning and further increases costs.

Whereas TDM is reaching its limits in terms of network elements and switching technologies, DWDM is just beginning. But as with any technology that is just beginning, obstacles will be in the way, and in the case of DWDM, the obstacles include management and performance impedance mismatches between networks. The International Telecommunication Union (ITU) has formed a study group that will look into the interoperability standards to ensure that traffic can move between vendor networks despite the underlying differences in the many different vendors' equipment.

Optical OADMs

Next-generation network services must be easily reconfigurable and they must support real-time provisioning. Demultiplexing all the wavelengths at each node is costly, it introduces delay, and it reduces the distance over which a signal can travel. OADMs, as shown in Figure 12.4, work much more inexpensively than demultiplexing all the wavelengths because they simplify the process they eliminate the costly electronics that are used to convert between light and electricity. Most OADMs use special filters to extract the wavelengths that need to be dropped off at a given location. For most vendors, the wavelength is fixed, so at the time of configuration, the carrier designates the individual wavelengths to be dropped at each location.

Figure 12.4. Optical add/drop multiplexing

graphics/12fig04.gif

Emerging DWDM applications address the growing desire for wavelength-on-demand. Individual wavelengths are assigned either to specific protocols, such as ATM or IP, or to specific customers. Alternatively, some providers might lease an entire dark fiber to each client, and the client would then purchase the customer premises equipment to route different protocols over each individual wavelength. This is opening the door to a whole new way of thinking about providing wavelengths to the long-haul carriers, to the MAN market, and to the customer.

The development of managed wavelength services depends on the development of wavelength changers and optical switches. A wavelength changer converts an optical signal to an electronic signal and then sends it to a laser that produces an optical signal at a different wavelength than the original. As you'll see in the next section, optical switches give carriers the capability to provision bandwidth automatically, instead of having to deploy technicians into the field. Optical switches also enable service providers to build mesh optical restoration, which gives them the flexibility of running different kinds of restoration in their networks. Finally, optical switches allow service providers to establish QoS levels associated with restoration.

Optical Switches

Optical switches, sometimes referred to as optical cross-connects or wavelength routers, are devices that reside at junction points in optical backbones and enable carriers to string together wavelengths to provide end-to-end connections (see Figure 12.5). They link any of several incoming lines to any of several outgoing lines and automatically reroute traffic when a network path fails. Optical switches are the optical version of the general-purpose switching system that provides flexibility and reliability in today's PSTN. Optical switches move transmissions between fiber segments and also enable some network management activities, including optical-layer restoration and reconfiguration, dynamic wavelength management, and automated optical-layer provisioning.

Figure 12.5. An example of an optical switch

graphics/12fig05.gif

There are three key issues in selecting optical switches:

         Number of ports Carriers are looking for devices that can scale to more than 1,000 ports.

         Automation Carriers want to provision strings of wavelengths from a remote console in real-time.

         Granularity Carriers want the switch to handle small as well as large bandwidths so that they can eliminate multiplexers.

First- and Next-Generation Optical Switches

Two types of optical switches are currently being produced: switches with electrical cores (that is, first-generation optical switches) and switches with optical cores. The electronics in first-generation switches slows their capability to work with the very high rates that the fiber itself can support. The future lies in the pure optical switches, but we still have to fully develop the microphotonics industry; thus, microphotonics is really the next revolutionary technology.

Optical switches fall into two main categories. First, the multiservice provisioning platform (MSPP) enables carriers to get a quick start on offering a full range of services. The MSPP resides either in the carrier's point of presence or at the customer site, and it incorporates DWDM, while also offering customers different grades of IP service, telephony, and other offerings. Second, big switches can be deployed at the carrier's local exchange. These switches act as on-ramps, funneling large volumes of traffic from IP and ATM backbones on and off the optical core.

Challenges in Deploying Optical Switches

Because we are in the early stages with optical switches, we have to deal with issues such as how to quickly provision services, how to accommodate billing, and how to elegantly separate services. In the next three to five years, we should start seeing these more sophisticated elements become available in pure optical form.

Again, with end-to-end optical networking, because transmission rates are reaching the Tbps, Pbps, and even the Ebps (that is, exabits per second) levels, the bottleneck is moving to the network elements. The faster the light pulses are emitted that is, the faster the data rates on the line get the more technically challenging it is to handle optical-electrical-optical conversions at line speed. Therefore, to fully take advantage of the capacity that's being created by WDM, fiber networks will need switches that are capable of rerouting light. The good news is that the cost of optical components has decreased by 40% in recent years, and it is expected to continue to drop by 40% to 60% per year.

The biggest problem that converged telcos are now facing is how to accurately forecast what their future bandwidth requirement will be. Transmission speeds are doubling every 12 months, so it is essential that we have infrastructures that are capable of providing a large amount of bandwidth on short notice and at a reasonable cost. Without intelligent optical networking, adding an OC-48 circuit over existing dark fiber can take between six and nine months. To automate provisioning, we need to also address how we can look into a wavelength to determine how to properly act on it.

Optical switches enable improved reliability, improved scalability, and flexible service provisioning. Another major benefit is that they reduce the capital required to add additional capacity, and the overall savings can then be passed on to the customer. Deploying optical networking technology in the metro area can bring the benefits of converged networks down to the customer's premises. The end-to-end optical infrastructure can then support advanced services such as true bandwidth-on-demand.

Optical Switching Fabrics

Optical switching fabrics, such as the following, provide subsystems that connect one wavelength to another:

         Microelectromechanical system (MEMS) switches A MEMS switch uses an array of microscopic mirrors to reflect light from an input port to an output port.

         Bubble switches Similarly to ink-jet printers, bubble switches use heat to create small bubbles in fluid channels that then reflect and direct light.

         Thermo-optical switches With thermo-optical switches, light passing through glass is heated up or cooled down by using electrical coils. The heat alters the refractive index of the glass, bending the light so that it enters one fiber or another.

         Liquid crystal display (LCD) switches LCDs use liquid to bend light.

         Tunable lasers Tunable lasers pump out light at different wavelengths, and they can switch from one wavelength to another very quickly.

Lucent's LambdaRouter is an example of a MEMS switch (see Figure 12.6). It switches lightwaves by using microscopic mirrors. One of these microscopic mirrors is small enough to fit through the eye of a needle. Hence, a LambdaRouter is essentially a switch with 256 ports by 256 ports in a one-square-inch piece of silicon. (Remember how it used to take a multistory building that took up an entire block to house a local exchange switch in the electromechanical era?) The LambdaRouter's 256 ports each start at 40Gbps, which means the LambdaRouter offers a total capacity of nearly 10Tbps. This device can intelligently switch or route the wavelengths without making any optoelectronic conversions. The lightwaves themselves tell the mirror what bend to make in order to route the light appropriately; they do this by using a digital wrapper, which is the equivalent of a packet header. Lucent's digital wrapper, WaveWrapper, is proprietary, but there are some movements afoot to standardize this intelligence, as discussed later in this chapter, in the section "Managing Optical Networks."

Figure 12.6. A MEMS switch

graphics/12fig06.gif

Xros, which was acquired by Nortel, is also focusing on the micromirror technology. The Xros device can scale to 1,024 x 1,024 wavelengths, and it fits in an assembly that's about 10 inches wide by 6 inches tall.

Another approach in optical switches, bubble switches, involves using ink-jet printer technology to switch packets in optical switches. In ink-jet printers, tiny enclosures are filled with gas, and they sit behind the ink in a printer. In front of each enclosure is a minute nozzle. When a character is called for, the gas behind the nozzles that form the letters is heated, and the ink is shot onto the paper. Agilent is applying this technology to optical switches, embedding tiny, liquid-filled cavities in switch fabric (see Figure 12.7). If a packet is supposed to stay on the same network, the liquid remains cool, and the packet passes through unscathed. If a switch is required, the liquid is heated and turns to gas. If the gas has the correct reflective properties and if the cavity is precisely positioned, the light is bounced in the right direction.

Figure 12.7. A bubble switch

graphics/12fig07.gif

Yet another optical switching approach involves thermo-optic switches. Light is passed through glass that is then heated up or cooled down with electrical coils. The heat alters the refractive index of the glass, bending the light so that it enters one fiber or another (see Figure 12.8).

Figure 12.8. A thermo-optic switch

graphics/12fig08.gif

An LCD switch, as shown in Figure 12.9, uses liquid to bend light.

Figure 12.9. An LCD switch

graphics/12fig09.gif

There are many issues to address with these evolving types of optical switches. Again, because we are in the early days of these technologies, we have not yet tested each of them, let alone under the full load of the network. We still have to address how well these technologies will scale. Carriers are looking for modular switches that can scale to thousands of ports. Switching speed is another consideration. Routing traffic packet-by-packet will require switching speeds that are measured in nanoseconds.

It is important to remember that we are in the early days. None of these switches have been tested in fully loaded environments, and undoubtedly some will prove to be more fanciful than fruitful.

Other Optical Components

The building blocks of optical networks, the optical components, are critical to the development of optical systems. In addition to the components described so far, the following are some important optical components:

         Lasers Provide light

         Gratings Single out specific wavelengths from a light source

         Filters Read incoming light signals

         Dispersion compensation modules Prevent smudging of light signals

         Variable attenuators Even out the strength of light signals in adjoining wavelengths

         Passive splitters Divert wavelengths into different fibers

         Amplifiers Strengthen attenuated signals

Key issues regarding the optical components part of the optical equation are cost and shortage. Currently, many components are extremely expensive. Also, many vendors have announced big investment in component production facilities. However, demand is likely to exceed supply for some components over the next couple years. Companies that do not yet have contracts for things like fiber-optic cables and fiber components may find themselves unable to obtain these items until the shortages are resolved.

To address the cost and shortage issues, vendors are developing ways of making optical integrated circuits, which are the optical equivalent of electrical integrated circuits. The goal is to consolidate large numbers of separate optical devices into a single chip, customizing them for different applications and drastically reducing costs and improving performance. A wide variety of materials are being used for different applications, including silica, polymer, and rare earths. Each has advantages and disadvantages in terms of the performance of the chip and the ease and cost of manufacturing it. Currently, vendors are producing relatively simple components, such as WDM chips, small switch modules, and passive splitters, but the future has a lot more in store in this arena.

 



Telecommunications Essentials
Telecommunications Essentials: The Complete Global Source for Communications Fundamentals, Data Networking and the Internet, and Next-Generation Networks
ISBN: 0201760320
EAN: 2147483647
Year: 2005
Pages: 84

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net