Coarse WDM comes of age for local carriers
Coarse WDM (CWDM) is a 4, 8, or 16 wavelength-per-fiber technology developed in 2000 and now widely used for metro networking over Gigabit Ethernet and Fibre Channel switching platforms. CWDM transceivers are available as gigabit interface converters (GBIC) and small-form-factor pluggable (SFP) components, two pluggable form factors commonly used in datacom switches. The use of pluggable optics enables wavelength selectivity by the end user at the time of deployment.
With CWDM technology, building a metro network is a simple plug-and-play exercise. Two different topologies—point-to-point and add/drop rings—can be implemented. In either case, substantial cost reductions are possible compared to non-WDM solutions because the same fibers can be used to carry more bandwidth. CWDM has been defined for metro network applications in which the addition of costly amplifiers is never required (see Fig. 1).
Many enterprise customers and competitive local exchange carriers (CLECs) have deployed Gigabit Ethernet networks around metropolitan areas. A typical node supported by one wavelength of Gigabit Ethernet consists of 24 to 200 ports of 100-Mbit/s fast Ethernet delivered over CAT5 wires pulled through a multi-tenant high rise or dropped around an enterprise campus. The 100-Mbit/s ports cost about $250 each including the following components:
- Network switches
- GBIC CWDM transceivers (Gbit/s data rates)
- Passive optical couplers—multiplexers or optical add/drop multiplexers (OADMs)
- Redundant optics for fail-over in the event of equipment or fiber problems.
The cost effectiveness of this type of network is dramatically better than non-WDM technologies. In spite of these advantages, local carriers have not adopted CWDM as readily as CLECs and enterprises. Part of the reason for this may be that many existing carriers and independent local exchange carriers (ILECs) purchased large quantities of equipment in the boom times of 1999 and 2000, when SONET and dense WDM (DWDM) were the solutions of choice. On a cost per delivered Mbit/s of service, those solutions are very expensive by current CWDM standards.
When more capacity is needed in a fiber network, carriers typically add additional wavelengths to the same fiber routes. In legacy SONET/SDH equipment, this is accomplished by ordering more blades for existing platforms. But in datacom equipment of the same vintage, pluggable optical components are now available.
CWDM emerged as a viable technology when the definitions for pluggable optical slots for current datacom equipment were finalized in 1996 for GBIC and in 2000 for SFP. It became possible to design a non-temperature-stabilized pluggable set of modules of wavelengths that met these specs and operated on less than 1.5 W. CWDM wavelengths cover the range from 1270 to 1610 nm in 20-nm increments. The most commonly used groups are from 1510 to 1570 nm for 4 wavelengths, 1470 to 1610 nm for 8 wavelengths, and 1310 to 1610 nm for 16 wavelengths. This spacing scheme was devised so that nonstabilized lasers would be able to move with changing operating temperatures but still stay "in band."
In comparison, DWDM devices typically require 10 to 15 W and multiple power-supply voltages to temperature-stabilize the laser wavelength and accurately hold it on the DWDM ITU-specified wavelength grid. Historically, DWDM devices required so much physical space that the technology could not fit physically into the pluggable slots in datacom devices.
CWDM transceivers plug into Ethernet and Fibre Channel switches available from all of today's most competitive infrastructure vendors. This is a simple and elegant process. CWDM does not require temperature stabilization for the laser components, and therefore uses less power. These devices were designed into existing datacom pluggable form factors and are installable in both legacy and future switches. Many fiberoptics vendors sell CWDM transceivers and passive components, and an ITU standard is nearing completion (see Fig. 2).
CWDM was never intended to retrofit into existing SONET/SDH platforms. Legacy SONET/SDH equipment vendors would need to redesign blades and entire systems to take advantage of the size, cost, and rapid-installation benefits that CWDM offers.
As the inherent advantages of CWDM help it gain in popularity, DWDM modules that provide higher channel densities are also becoming available in a GBIC-pluggable format. These can be mixed with CWDM devices to add more capacity on the same fiber.
Transceivers with 100-GHz spacing in a GBIC form factor are available, enabling larger channel-count applications to be met with the same datacom technology, ease of installation and maintenance, and compatibility with existing field hardware. DWDM GBICs should be available in 50-GHz ITU spacing in early 2003, with each transceiver having a limited wavelength tuning range to keep inventory levels lower and implementation simpler for end-users.
CWDM wavelength bands are a series of 8- or 16-wavelength swathes of optical spectrum. When CWDM is deployed, a customer can later decide to convert or upgrade one or more of the CWDM bands from its single unstabilized wavelength to higher-capacity DWDM traffic instead. For example, an 8-wavelength CWDM system can carry up to 32 wavelengths per band at 50-GHz spacing.
In a metro system, this means that a simple, low-cost network can be established with six to eight wavelengths (one wavelength per node) deployed on a ring around a city or campus. As buildings and metro regions experience traffic growth and need additional services, the CWDM transceivers can be replaced by DWDM gear to expand only those nodes that need more channels while leaving the rest of the system unmodified. Using this upgrade path, hundreds of wavelengths become possible in a system that started out as a cost-effective solution for just a few wavelengths (see Fig. 3).
The cost benefits, flexibility, and ease of implementation offered by CWDM and pluggable DWDM components make these two technologies very viable for ILECs, CLECs, and enterprise customers to implement new networks and expand existing networks. We expect this trend to gather considerably more steam during the next 6 to 12 months.
Frank Levinson is founder, CTO, and chairman of Finisar, 1308 Moffett Park Dr., Sunnyvale, CA 94089-1133. He can be reached at firstname.lastname@example.org.