Users of LAN data services have come to expect continual advancements in network speeds. IT professionals also know to expect the price of a given generation of networking equipment to decline as time goes by. These developments have been predicted by Moore’s Law. The challenge is to make sure that the LAN infrastructure can cost-effectively allow the IT professional to upgrade to the speeds expected and demanded by users.
Clearly, effective planning goes a long way toward meeting this challenge. One way to approach the planning process is to “grade” optical cabling options according to fiber type and the transmission distance each type supports for a given data rate. By matching the capabilities of each fiber grade with the speeds and reach each part of the LAN is expected to support over its lifetime-not to mention the project’s budget-network planners can meet their anticipated requirements cost-effectively, without paying for capabilities they aren’t likely to need.
Ethernet is the dominant protocol in the LAN. History tells us that every five years, the speed of Ethernet will increase by a factor of 10. In 1985, 10Base-T was released, followed by 100Base-T in 1992. 1000Base-X (fiber) was released in 1997 and 1000Base-T was released in 2000. 10GBase-X (fiber) was released in 2002, and 10GBase-T is targeted for release in 2006. Meanwhile, 40-Gigabit Ethernet is expected to be the next step. In each case, the new technology was relatively expensive when first released and purchased only by customers that absolutely needed the improved data rate. Eventually prices declined, and the data rates that once seemed excessive became commonplace.
Increased data rates lead to a more efficient cost of data, upon which almost all businesses depend. The business that uses its resources most efficiently has a cost advantage over its competition that typically shows up in the form of improved customer service.
Users usually notice the results of the installation of new hardware that supports a step up in Ethernet generations. In contrast, the cabling infrastructure is hidden from most users (both figuratively and literally). The average user of data services cannot “see” the difference in an improved cabling infrastructure. Thus, they are not concerned about the medium used to transport the information as long as that information gets to them faster. Most IT professionals know the difference that cabling can make in the performance of the network; however, they also know it is much more time-consuming and inconvenient to change the cable infrastructure than to change the network equipment.
Fortunately, the Telecommunications Industries Association (TIA) has set a benchmark that cabling infrastructure should be designed to last at least 10 years. With proper planning, infrastructure can be designed to support growing data rates during that 10-year life expectancy. Based on the Ethernet history just discussed, that means two future generations of Ethernet, which may get five years of use each (see Figure 1).For example, 100Base-T is still a popular protocol in the horizontal part of the LAN. If new cabling is being installed, it is a good rule of thumb to design the infrastructure to handle at least up to 10GBase-T (assuming approximately five years to get to 1000Base-T and another five years to get to 10GBase-T). Today, a typical backbone has either a 100-Mbit/sec or 1-Gbit/sec data rate, which would lead to an expectation of 10 Gbits/sec or 40 Gbits/sec in 10 years.
With the advent of 1000Base-X (1997), the bandwidth of the popular FDDI 62.5-µm multimode fiber (MMF) has become a limitation. Partially, that’s because the fiber was optimized for use at the 1300-nm (long) wavelength, while newer high-speed technologies use the 850-nm (short) wavelength. Also, the change from LEDs to lasers showed some operating deficiencies in these traditional fiber types.
These changes led to a wave of new MMF types and standards-based test and measurement methods. The most recent change to the standards is the use of a differential modal delay measurement to predict the effective modal bandwidth for laser-optimized MMF. This new approach provides assurance that the bandwidth of the fiber is sufficient for these extremely high data rates.
Proper planning should allow the use of short-wavelength electronics whenever possible. In general, the cost of optoelectronic devices is significantly less expensive at the short wavelength than the long wavelength. However, the long-wavelength option is useful because it can be easier to implement on the installed base. With the various fiber types and wavelength options, how can you tell which combinations are the most cost-effective?
The IEEE has the responsibility of defining the operating characteristics of Ethernet equipment. The IEEE standards for 1-Gigabit Ethernet (GbE) and 10-GbE define the distance limitations for the various fiber types. Since fiber cables are often used for their ability to transmit signals over longer distances, the distance limitation by application becomes an important criterion to consider in the design and choice of the appropriate fiber cabling system (see Table 1).
We can use the data in Table 1 to create a grading system that we’ll call the Multimode Fiber Grade Selector. The grading system categorizes the fibers into six groups, based on bandwidth and application support: Grades 1, 2, and 3 are 62.5/125-µm MMFs and Grades 4, 5, and 6 are 50/125-µm MMFs.
Grade 1 is the standard FDDI-grade 62.5/125-µm fiber specified in TIA/EIA-568B.3. Grade 2 provides improved laser performance at the short wavelength. Grade 3 is an enhanced 62.5/125-µm fiber with a specific length benefit for 1000Base-SX networks. Grade 4 is the “standard” 50/125-µm multimode specified in TIA-568B.3. Grade 5 is the laser-optimized 50/125-µm fiber specified in TIA-568B.3-1. Grade 6 is a proprietary enhanced laser-optimized fiber.
Table 2 compares the short-wavelength 10-GbE (10GBase-SR) distances for each of the fiber grades along with their relative costs.
It should be noted that Grade 6 is a fiber not recognized by standards at this time. However, it offers greater bandwidth than other fiber grades and can be used to create engineered links that offer a longer link length than other fibers can support. For example, based on the 10-GbE model, as the bandwidth of the fiber increases, the power penalty decreases. The power savings provided by Grade 6 fiber can be reallocated to the cable insertion loss to create a greater loss budget, which can translate into longer link length.
Alternatively, it can support links within the 300-m limit, in which additional connector loss must be accommodated. As an example, pre-terminated fiber installations have become a popular way to install fiber backbones. The typical method involves having a trunk cable terminated at each end with a 12-fiber MPO ribbon connector. A cassette is installed at each end of the trunk cable. The cassette has the MPO connector at one end and 12 discrete connectors (e.g., LC connectors) at the other end. The link now needs to account for two additional connections-the MPO transitions from the trunk cable to the cassettes. The additional power margin gained by the use of Grade 6 fiber will allow the extra 1.5 dB needed by the MPO connections.
All optical communications require the use of a transceiver. This device takes the electrical signal generated by the network physical layer and drives a light source that is used as the source for the optical link. As has always been the case, electronics such as the transceiver are the largest contributor to the cost of an optical link.
Once requirements reach 1000Base-X, the user has a choice of transmission wavelengths and therefore transceivers. When specifying a GbE switch, the transceiver can be short-wavelength (1000Base-SX, 850 nm) or long-wavelength (1000Base-LX, 1300 nm). The most important contributing factor in the choice is the distance that needs to be covered. As shown in Table 1, the wavelength and fiber chosen combine to provide a minimum supported link length.
However, cost is also a significant factor. 1000Base-SX electronics are significantly less expensive than 1000Base-LX. At this time, the short-wavelength electronics are two to three times less expensive.
With 10-GbE, there are three options for LAN applications: short-wavelength (10GBase-SR, 850 nm), long-wavelength for singlemode fiber (SMF) only (10GBase-LR, 1310 nm), and long-wavelength for multimode or singlemode fibers (10GBase-LX4, 1310 nm WWDM). At this time, there is about a 500% price difference between 10GBase-SR and 10GBase-LX4, where the long-wavelength option is more costly.
Since the network can use a variety of wavelengths and a variety of fibers, the number of possible implementations becomes relatively large. To make the best choice from all of the available options, the IT professional must look at how these costs and performances combine to provide the most cost-effective solution.
To illustrate how the Multimode Fiber Grade Selector can be put into practice, data was compiled to determine the relative cost of installing a new fiber-backbone network in two scenarios. The various MMF types were compared along with SMF and a hybrid cable containing both multimode and singlemode fibers. Equipment prices were gathered from various retail sources and an average price was used in the model. The labor costs were gathered by polling both union and non-union contractors.
The first of the two scenarios was a 15-story building backbone. Specifications included: 12-fiber cable installed between closets, all 12 fibers terminated in patch panels (24 fibers in the hybrid design), two active links per cable (four active fibers/eight spare fibers), and short-wavelength optics used where possible.
The building backbone design called for a 12-fiber cable to be installed and terminated between a main crossconnect in the basement and a telecommunications closet on each of the 15 floors. In the case of the hybrid cable, 24 fibers were said to be terminated; 12 multimode and 12 singlemode. Each cable had two active links (four active fibers) to simulate two switches in each closet. The cost of the switch was not included in the cost calculations since that would be considered part of the horizontal cabling.
The design called for the initial backbone data rate to be GbE, upgraded to 10-GbE in five years. The 10-Gbit/sec equipment-price declines were predicted based on Moore’s Law.
The second scenario comprised a nine-building outside plant campus backbone. Its specifications included all backbone distances at 500 m, 12-fiber cable installed between closets, all 12 fibers terminated in patch panels (24 fibers in the hybrid design), two active links per cable (four active fibers/eight spare fibers), and short-wavelength optics used where possible.
When TIA-EIA-568A was revised to TIA/EIA-568B.1, the 500-m intermediate backbone distance became 300 m. However, in some instances, 500 m is still the actual distance. For our cost-comparison study, the campus model used the same set of assumptions as the building backbone, except that all eight of the cable lengths were 500 m.
In the first scenario, although the cost of the various grades of MMF varies significantly, the rest of the materials (connectors, patch panels, adapters, etc.) and the labor are identical. As shown in Figure 2a, the installed costs of Grades 1-4, along with singlemode, are virtually identical. The cost of the 1-Gbit/sec electronics drove up the cost of the singlemode option. When the 10-Gbit/sec electronics were added, the advantage of using Grade 5 or 6 fiber became clear. Even though the cable comes at a significant cost premium, the savings in electronics was dramatic.
For the second scenario, the cost of the various grades of multimode fiber also varied significantly and the rest of the materials (connectors, patch panels, adapters, etc.) and the labor were identical. As shown in Figure 2b, the cost premium associated with the various fiber types becomes more evident, since the amount of cable relative to the rest of the costs increased. However, the cost of the 1-Gbit/sec electronics still drove up the cost of the singlemode option.
When the time comes to upgrade to 10 Gbits/sec, the 500-m length would prevent the use of Grades 1-5 fiber. Only the Grade 6 option would allow the use of the lower-cost 10GBase-SR modules.
Proper planning of a structured cable plant can provide the opportunity for significant cost savings while allowing an upgrade path for future network speed upgrades. The use of Grades 5 and 6 fibers provide the most cost-effective solution for the 10-year life expectation of backbone cable plant. Grades 5 and 6 provide the maximum bandwidth at the short wavelength that coincides with the lowest-cost optoelectronic equipment.
Grade 5 meets the standards requirements of IEEE 802.3ae and TIA/EIA-568-B.3.1. These standards were created to allow the enterprise users to maintain a cost-effective multimode cable plant with an upgrade path to higher-speed technologies.
Grade 6 provides significantly higher bandwidth, which provides additional operating margin to be used for added cabling distances or additional connections. That can be an important factor for plug and play installations. While Grade 6 is outside the definition of the standards, its proper use can save a significant amount of money. It is critical that proper documentation be maintained when using Grade 6 fiber so that the engineered link values can be referred to as needed in the future.Michael Connaughton, RCDD, is fiber-optic sales manager at Mohawk, a Division of Belden CDT (Leominster, MA).