Analysis of markets for fiber optic network bandwidth suggests that the investment boom has by no means played itself out. Between 1998 and 2000, new submarine cables increased Trans-Pacific bandwidth from 14 Gbps to 244 Gbps. That achievement pales in comparison with what is yet to come. Once fully upgraded, cables now under construction could add an astounding 17 Terabits per second (17,000 Gbps) to the Trans-Pacific route. Terrestrial networks boast even greater upgrade ca pabilities. A handful of pan-European networks hold the potential to carry 1 Petabit per second (1,000,000 Gbps) of traffic between cities, equivalent to 2 Mbps for every person in Europe. Because operators can upgrade these new systems gradually, the effects from the huge capital infusions during the late 1990s will not be fully felt for many years to come.

A booming stock market enabled financing for multiple network build-outs in the late 1990s. However, it was technology that fueled the greatest increases. Whereas early fiber optic systems used a single wavelength to transport information, modern networks employ Dense Wavelength Division Multiplexing (DWDM) to make use of many different light frequencies within a cable. Such advances mean that, with all wavelengths lit, bandwidth on a single network currently under construction could dwarf the combined existing capacity of all the other cables on the same route.

Despite the well-publicized problems of some builders, many are still adequately funded. Global Crossing, Interoute, Level 3, Telia and TyCom all appear to have ample resources to continue their builds, although more conservative financial management may slow the pace of deployment.

Compared to the astounding increases in lit and potential bandwidth, the actual market structure of the industry has remained relatively stable in the last twelve months. No major new supplier has emerged since TyCom’s entry into the bandwidth provisioning market in early 2000. However, many capacity providers are moving into related markets. Companies that initially supplied only wholesale bandwidth have added new services such as colocation, IP transit, switched minutes, last mile facilities and content distribution.

The massive upsurge of bandwidth in many parts of the world has sparked extraordinary price declines; it also has engendered a bewildering array of new products for buyers.

Optical Revolutions

DWDM, the key development in fiber optic technology of the 1990s, will continue to wring more bandwidth from a single pair of optical fibers over the next several years. Other emerging technologies continue to widen the boundaries of the possible. In 2001, some suppliers will deploy networks that exploit a new optical "window" at 1,600 nanometers, the so-called L-band. By 2002, the first 40 Gbps wavelength system should be in place. Meanwhile, improvements in both laser and fiber technology will enable designers to squeeze more signals into each window. Current networks space wavelengths 100 or 200 GHz apart; those to be deployed this year and the next will space wavelengths 50 or even 25 GHz apart. Ultimately, these new technologies will allow operators to transmit up to 320 wavelengths on a single fiber pair.

Just as DWDM revolutionized fiber optics in the 1990s, optical switching could dominate the new decade. Switching tasks cannot easily be accomplished in the optical domain because light signals and photons are much more difficult to manipulate than electrical signals and electrons. Hence, at every point in a network where signals must be separated and routed discretely – that is, at every single node – the light signal must be converted into an electrical signal, processed and then converted back into an optical signal. This continual optoelectronic conversion comes at a price. 

Each nodal point requires a complete set of multiplexing and demultiplexing equipment, which is very expensive – once purchased, it must be housed, maintained, managed and powered. Moreover, traditional technologies were not designed to handle the very rapid increases in bandwidth or the continual reconfiguration of networks brought on by the Internet. Every time a new wavelength is required, skilled network designers must study network configuration maps, assign paths and send out technicians to manually set up connections at every site. A shift to all-optical networks could introduce vast savings in both time and money; it also could enable new bandwidth products featuring rapid provisioning and network flexibility – a far cry from the often-cumbersome arrangements still in place.

While construction costs dominate initial expenses, especially in terrestrial networks, the high cost of DWDM equipment accounts for a larger proportion of ongoing costs. Optimistically, one can expect the costs of electronic equipment to fall with improvements in the price-performance ratio of electronic componentry. DWDM and related technologies will likely continue to reduce the costs of bandwidth for at least the next two to three years. The main technological challenges will emerge from long-standing difficulties in applying optical techniques to network management, network architecture, switching and provisioning.