An introduction to optical networking
Jeff Hecht,Contributing Editor
Although the ultimate goal of an all-optical network has yet to be realized, the use of optics in network management promises significant increases in capacity and versatility.
Optical communications used to involve little more than the transmission of signals from point to point over an optical fiber. More recently, however, the field has evolved to the point that not only can optical networks transmit signals optically, but they can also organize and process them in optical form. The benefits include networks with increased capacity, speed, and versatility. In this new series, contributing editor Jeff Hecht takes an "end-to-end" look at optical networks, explaining the technologies involved and highlighting the many challenges to be addressed in search of the ultimate goal: the all-optical network.
Optical networking may sound like just another fancy buzzword that merely repackages fiberoptic communications so customers and investors think they're getting something new. However, optical networking does mark a fundamental shift in the architecture of fiber communications. Early fiberoptic systems were merely transmission pipelines, carrying signals from point to point. Optical networks organize and process signals in optical form as well as transmit them optically, expanding the realm of light. The ultimate goal is marked by another vague-sounding buzzword, the all-optical network, in which signals stay in optical form throughout the network.
The real allure of optical networking is as a new tool for coping with a rising volume of telecommunications traffic that is outstripping the capacity of old technologies. Optics offers a whole new layer of organizational capabilities. Signals that were electronically organized by data rates can be optically organized by wavelengths—optical channels or lambdas. A look at the recent evolution of telecommunications will help one understand how optical networking opens new possibilities.
PIPES AND SWITCHES
The roots of today's digital telephone network date back to the 1960s, when telephone planners set up a hierarchy of standard data rates based on time-division multiplexing of digitized signals. As the volume of telephone traffic increased, planners added a family of higher data rates (see Fig. 1).
The signals are organized and processed electronically as interleaved streams of bits to make faster signals. The old Bell System and the International Telecommunications Union originally set different series of standard rates starting with individual digitized voice channels at 56 or 64 kbit/s, and topping out at a few hundred megabits per second. The advent of single-mode fiberoptics allowed much faster data rates, leading to the family of synchronous optical network (SONET), or synchronous digital hierarchy (SDH), data rates also shown in Fig. 1. Today, the top data rates in practical use are 2.5 or 10 Gbit/s, although a few manufacturers have begun to offer 40-Gbit/s systems. In the near term, practical problems will limit 40-Gbit/s transmission to relatively short distances.
The primary role of optical fibers has long been to transmit high-speed bit streams from point to point, between nodes in the network. Electronics at the nodes then process and switch the signals, multiplexing or demultiplexing them to different data rates, and directing them to different nodes. In essence one can view the telecommunications network as built of two main components—pipes, which transmit signals, and switches, which process and direct the signals. Fibers have been the pipes for high-speed signals. In optical networks, they are becoming the switches as well.
The first big step toward optical networking was the advent of wavelength-division multiplexing (WDM). As far back as the late 1970s, engineers realized that optical fibers could carry separate signals at different wavelengths. In the early 1980s, the old Bell System used two WDM channels when it built the first long-distance fiber system between Boston and Washington using multimode fiber. However, that system required separate electro-optical repeaters for each channel approximately every 8 km. Single-mode fiber systems carrying higher data rates proved far more economical. WDM reappeared once erbium-doped fiber amplifiers were developed, because the optical amplifiers could simultaneously amplify signals at multiple wavelengths.
The initial attraction of WDM was its ability to multiply the capacity of a single fiber. Instead of carrying a single time-division multiplexed (TDM) channel at 2.5 or 10 Gbit/s, a fiber could carry 4, 8, 16, 32, 40 or more optical channels at different wavelengths, each at the same data rate. At first, engineers thought of this as multiplying the TDM rate. For example, a 16-channel, 2.5-Gbit/s system was considered equivalent to a 40-Gbit/s system.
However, the WDM signal consists of 16 TDM bit streams rather than one, and that makes an important difference to telecommunications carriers. Remember that time-division multiplexing combines many different signals to transmit in one unified, higher-speed signal. The advantage is that transmitting one high-speed signal costs much less than transmitting many signals at slower data rates. The disadvantage is the need to keep track of all the different components of the combined signal and demultiplex them when they reach their destination. The original telephone industry formats interleave signals so thoroughly that they must be broken down completely, even if one only wants to extract a single telephone signal. The newer SONET formats offer more flexibility in breaking out signals, but demultiplexing remains an issue because it requires electronics to decode and direct the signal.
Wavelength-division multiplexing makes signal manipulation possible on the wavelength level. Combining four 2.5-Gbit/s signals into a single 10-Gbit/s data stream requires an expensive electronic TDM multiplexer and a 10-Gbit/s optical transmitter—plus a corresponding receiver and demultiplexers to pick out one of the signals. If the four 2.5-Gbit/s signals are sent on separate optical channels, a filter can pick off the desired optical channel without disturbing the rest of the channels. The same principle applies in any WDM system: optical components can separate one optical channel from the combined signal without electronics.
From the standpoint of a telecommunications carrier, this gives the signal increased granularity by allowing the isolation of smaller chunks of the combined signal. Granularity is important in managing signal transmission and services. Few customers may need 2.5 Gbit/s today, but the aggregate of signals being transmitted between pairs of points can easily reach that level. For example, a carrier might group signals sent from San Francisco into several different optical channels on the same fiber. Separate add/drop multiplexers could drop one channel at San Mateo, another at Palo Alto, a third at Sunnyvale, and carry three more all the way to San Jose (see Fig. 2).
Emerging regional or "metro" optical networks may assign separate optical channels on the same fiber to customers who can use those channels to transmit in any format they choose. One might use Gigabit Ethernet, another might transmit raw digital video, and a third might use 2.5-Gbit/s SONET. These optical channels become separate information pipelines through the same fiber, which the network operator can manage regardless of content. New, highly automated systems speed provisioning, the process of finding transmission capacity and assigning it to customers, which is a major concern to carriers. In short, the optical network promises carriers a better way to manage their growing volume of traffic.
OPTICAL SWITCHING AND ROUTING
Implementing optical networking poses some serious challenges. Fiberoptic "pipes" are well-developed, but optical switching and signal management is in its infancy. Emerging system designs pose some diverse requirements.
Ways to switch both individual optical channels and combined signals on various time scales are obviously needed. Both cases can be important. Today it may suffice for an add/drop multiplexer to always drop a certain channel in San Mateo, but eventually carriers would like to dynamically reconfigure their systems. Operators of metro networks want to be able to reroute signals quickly so they can supply customers with new capacity as soon as the order arrives. The ability to switch whole fibers full of signals is vital for route restoration in case traffic has to be rerouted around a cable break.
Optical networking also would benefit from wavelength converters, which could shift a signal from one wavelength to another (see Fig. 3). Wavelength conversion promises carriers the flexibility to use all the available optical channels, instead of just those that happen to be open along the entire transmission path. For example, a carrier might not have a single wavelength available all the way from Cleveland to Philadelphia, but might have 1540 nm from Cleveland to Pittsburgh, 1542 nm from Pittsburgh to Harrisburg, and 1538 nm from Harrisburg to Philadelphia.
Many details remain in implementing such systems. The easiest way to convert wavelengths today is electro-optically, by converting the input optical signal to electronic form and using it to drive a transmitter at the desired wavelength. Developers are exploring other alternatives, such as all-optical devices based on nonlinear optics or optical amplifiers, and driving tunable lasers that can select any available optical channel.