Expansion alternatives: WDM, TDM, fiber
Several application factors can affect both initial and long-term expansion costs.
BY JEFF D. MONTGOMERY
There is broad industry agreement that the global volume of data transported over fiber will continue its rapid expansion, though at varying year-to-year growth rates. The arguments are not whether, but how rapidly, how soon, and how that will differ relative to location in the network structure. A corollary to the forecasting of this growth is the question of how it will be accommodated. There are three basic alternatives:
- WDM, which produces more channels on one fiber by modulating multiple channels, each on a separate, isolated wavelength.
- TDM, which involves modulating existing channels at higher data rates (typically in 4x steps).
- Install more fibers, parallel to existing fibers, with WDM and TDM levels comparable to currently lit fibers.
Each of the three alternatives comes with its own cost assumptions and dynamics. For example, fiber deployment cost increases slowly, relative to capacity; the deployed cost of a 256-fiber cable may be only 10% more than a 24-fiber cable. On the other hand, WDM cost increases linearly with capacity.
TDM cost trends are more complex than those of the other two alternatives. They depend on the maturity of the components. Early-production 10-Gbit/sec transmitters may cost 1.5x the cost of four 2.5-Gbit/sec devices, while mature production may reduce that to 0.5x. The four 2.5-Gbit/sec solution also requires four fibers or four WDM filter channels. Higher data rates, 10 Gbits/sec and up, require expensive grooming components such as chromatic and polarization mode remotely tunable dispersion compensators and/or new, more expensive fiber optimized for higher data rates.
It also should be noted that the cost of fiber extends beyond the fiber itself. There are three significant cost elements of fiber deployment: fiber cable, installation labor, and right-of-way cost.
After the network cable is deployed and operating, further increases of capacity generally are most economically made by a combination of WDM and TDM. Either of these can accomplish incremental increases at moderate cost.
A key factor in the relative trends of deployment of fiber versus TDM versus WDM is the ability of these alternatives to scale and the cost of that scaling. Of the three alternatives, TDM is most limited.
Unwilling to be blindsided, as many were at 10 Gbits/sec, several vendors are now moving 40-Gbit/sec components and equipment toward commercial offering. However, the challenges of moving 40 Gbits/sec down an optical-fiber channel are much greater than at 10 Gbits/sec (16x for chromatic dispersion; very challenging for polarization-mode dispersion compensation) and more expensive to solve.
It appears likely that 40 Gbits/sec will be practical for new networks, with a new-formulation fiber and other new components, but is economically questionable for upgrading existing long-haul networks.
Deploying fiber across the open countryside is a virtually unlimited expansion option. Downtown, the situation is a lot more difficult and complex. Many older cities have downtown underground utility tunnels that can accept substantial additional fiber cable. In the newer cities, however, and in the suburbs of older cities, new trenching often is required for each new cable, making new fiber a less attractive expansion option.
That leaves WDM as the bandwidth expansion option with relatively unlimited upside potential. Currently, there are very few network spans with more than 32-wavelength DWDM filters installed, and these typically have a substantial share of their channels still dark. Of the lit channels, many are carrying less than 10 Gbits/sec. The 32-wavelength DWDM is nearly all in long-haul networks; metro/access networks are typically four to 16 wavelengths, and many fibers have no WDM. So with currently available products, there is room to increase existing long-haul throughput by two to three orders of magnitude by DWDM and lighting all channels. Metro/access has even greater WDM upward expansion room.
Metro/access links, and especially access, are now recognized as the "missing link" that is retarding the expansion of both Internet Ethernet and telecom SONET/SDH traffic. Regulated telcos in the United States have been reluctant to install fiber to subscribers, since competitive local-exchange carriers might also use the fiber. However, this attitude has changed significantly over the past 18 months, and telco access fiber connections grew impressively in 2001. On the Internet side, new service providers such as Cogent Communications are making major subscriber connection progress.
These thrusts bring a focus on the WDM/TDM/ fiber choice for access. The fiber now deployed for access in the United States is optimized for the 1310-nm band. To date, the only 1310-nm singlemode transmitters commercially available have been SONET; OC-3, -12 and -48 are widely deployed. As 1310-nm Gigabit Ethernet singlemode VCSEL-based transceivers phase into commercial volume in 2002, Internet service providers will switch to them.
There is major contention between DWDM (1.6-nm spacings or less) and coarse WDM (CWDM-2-nm spacings or wider) for access applications. CWDM filters, in comparable volume, will cost less than those for DWDM. More importantly, substantially lower-cost free-running transmitters versus precise wavelength-locked transmitters, can be used with CWDM. The downside is CWDM has much less upward scalability. The WDM/TDM/fiber tradeoff is different in the access from long-haul; the TDM (data rate) of the transmitter typically is set by the subscribers' level-of-service choice, rather than by an economics tradeoff decision at the central-office digital-crossconnect-switch interface. So the service provider's choice focuses on WDM (coarse or dense) versus more fiber.
The fiber basic cost is relatively low, with runs averaging less than 10 km for subscriber-to-point of presence (or central-office) links. That equates to about $1,200-$1,500. Installation and right-of-way costs, however, can vary from moderate to astronomical, depending on availability of underground duct and/or overhead poles and other factors.
Jeff D. Montgomery is chairman and founder of ElectroniCast Corp. (San Mateo, CA), an optical technology market-research and analysis firm.