The fundamental advantages of photons, or light pulses, as carriers of digital data have made optical-fiber communications networks the dominant component of the worldwide communications infrastructure in general, and of its inner layers in particular. The fact that the data is carried by photons means it can be transmitted at very high bit rates, and the fact that the photons are trapped within the medium of optical fiber makes it possible to avoid the crosstalk that is typical of coaxial cable, the electronic counterpart of optical fiber.
Dense wavelength-division multiplexing (DWDM), which combines multiple wavelengths of light into a single stream, taps an additional inherent advantage of photons. Exploiting the fact that photons of different wavelengths do not normally interact enables us to transmit many channels of data in parallel in the same fiber by using photons of a different wavelength for each channel. DWDM reigns as the leading technology for transmitting high volumes of data traffic over long distances. DWDM systems with 80 channels per fiber are currently being deployed, and systems with 160 channels/fiber are expected to be available by the end of the year.
The use of optics in networks today is limited, however, to transmitting data from one point to another. The data generated as an electronic signal is first multiplexed in the time domain, interleaving bits from various signal streams, to become a very fast string of bits carried by an electric current. It is then converted into optics and becomes a single-wavelength channel, a flow of photons at a specified wavelength.
Many such channels are combined in the wavelength domain, becoming the DWDM signal that is sent into the fiber. However, at the end point of the fiber the data is converted back to electronics, since the systems that route and switch the data channels are electronic. Thus, although the strength of optics is fully exploited in the single point-to-point segments of the networks, the overall architecture of the network remains electronic in nature.
A diagram of a DWDM network can easily be adapted to represent a 1970s-vintage network. By taking an unaltered network diagram and replacing the DWDM labels with appropriate labels from the 1970s, such as "coaxial cable" for "optical-fiber cable," a vintage network diagram can be created without changing the network architecture.
Expedient vs. optimal
The optical networks that are currently being deployed, while masquerading as "high-tech" or even "all-optical," are merely implementations of 1970s-vintage networks utilizing optical components. While this is an expedient approach to building an optical network, it is clearly suboptimal.
The capability of the long-haul segments of the network to carry large volumes of data traffic is growing exponentially, to meet the ongoing growth in the demand for bandwidth by the expanding circle of Internet users. At the same time, the performance growth rate of the electronic switching systems lags substantially behind. Thus, barring a tremendous leap in the performance of silicon, it is clear that keeping up with the Internet will soon require switching devices that simply cannot be built with conventional technology.
To meet that challenge, an intensive R&D effort has been devoted in recent years to developing optical-switching technologies. However, most of the optical switches that are being developed do not distinguish among the single-wavelength channels of the DWDM signal. Consequently, those switches can handle one channel of data per port, and in that respect they are similar to their electronic counterparts.
Classical switching systems, whether mechanical, electromechanical or electronic, are orthogonal switching systems. Since they are established upon a basic premise that no two signals can occupy the same space at the same time, they require that one, and only one, switch element in a given row and column be active at any one time. This ensures that two signals are not allowed to occupy the same space at the same time.
However, this same insurance strikes at the heart of the optical network. It is precisely the capability of two optical signals, on different wavelengths, to occupy the same space at the same time that makes the DWDM point-to-point segments extraordinarily powerful. That principle must now be expanded to the switching arena so that its strength will be fully exploited by the network as a whole.
When building a switching matrix based upon a crossbar paradigm, the number of switch elements required is n2. The crossbar switching paradigm, an orthogonal paradigm in which only one element per row and column can be activated, quickly grows beyond manageable numbers. As an example, a switching system to handle 16 inputs to be delivered to four outputs will require 162 or 256 switching elements and a demultiplexer external to the switching fabric. Increase the number of inputs to 256, and the required number of switching elements skyrockets to 65,536.
In an attempt to lower the number of physical elements involved in deploying such a switch, a few creative technologists have developed the capability of moving a single physical element in three dimensions, to act as multiple switching elements. This dramatically compounds the complexity of the fabric, of course, while the number of switch points remains constant.
For a switching system that is capable of switching 1,024 wavelengths, the required number of switching elements explodes to 1,048,576. A system capable of switching 2,048 wavelengths requires a staggering 4,194,304 switch points. While a 2,048-wave-length switch sounds extraordinary, it is important to note that such a switching system would be capable of managing only 12 optical fibers with 160 wavelengths per fiber, or six optical fibers with 320 wavelengths/fiber.
Since the fastest area of growth in the DWDM network is the wavelength domain, it quickly becomes clear that the ability to build switching systems capable of managing thousands of wavelengths is a near-term requirement. As DWDM networks grow in fiber and wavelength count, they will quickly overwhelm the ability of even the most elegant orthogonal switching systems.
The extraordinary power of the photon makes the impossible trivial. A powerful and compelling example of this is something we call Trellis Algebra, a fundamental shift in switching paradigms. By breaking the orthogonal mind-set that has dominated switching fabrics to date, an entirely new, and much more scalable, architecture can be created.
Trellis Algebra is based on the electroholographic optical switch, a wavelength-selective switch developed by Trellis Photonics. An array of electroholographic single-wavelength switching devices forms the optical module that is the basic building block in Trellis' Intelligent Lambda Switch.
In the optical module, all signals destined for an output port are sent to the same physical fiber; that is, remultiplexing occurs in the switching fabric itself. Thus, since there is no separate remultiplexing required after the switching function, multiple, separate switching fabrics can be coupled in a single-stage, nonblocking architecture.
The Trellis Algebra concept, which cannot be implemented by an orthogonal switching fabric, can be utilized to achieve a large switching architecture or it can implement a "pay-as-you-grow" vehicle to enable the service provider to start with a relatively small switching fabric and add parallel switching fabrics as the need arises.
In addition, this approach provides an ease of maintenance that is not possible with monolithic switching architectures, since maintenance can be performed on each switching matrix without disrupting the performance of the rest of the switching fabric.
History and experience often overwhelm the better judgment of the best of engineers. As the optical-communications systems being deployed today are merely extensions of their electrical forebears, we continue to design optical systems based upon electronic rules, regardless of their applicability.
Only by constantly questioning the basis for the rules for system architectures can we ensure that we access all of the power that lies within a given technology. The semiconductor taught us that transistors are designed to be wasted: a revolutionary thought at the time. Trellis Algebra teaches us a new basic truth: Optics are optics and not electronics, and they enable new and astonishing capabilities that we have not yet even conceived.