Cost, reliability and upkeep have long been the barriers preventing network owners from choosing all-fiber networks. Here, a fiber advocate gives his take on the trends that are breaking down these walls and making 'fiber to the desk' a viable option.
March 1, 2000
Fiber to the desk (FTTD) has been a topic of conversation in the cabling industry for several years. Yet this potentially huge market has been an elusive quarry with only a small percentage of networks actually installing FTTD.
There have been some valid issues that have limited fiber in the horizontal, but changes within the industry are now reducing or eliminating these barriers.
The growing demand for speed and bandwidth has made Gigabit Ethernet (GbE) over optical fiber increasingly popular in backbone and riser links. This popularity also is pushing GbE — and fiber — steadily towards the desktop. GbE to the desktop may seem far-fetched, but Apple already offers a GbE card as an option for its G3 computer and both Phobus and Agilent produce GbE network interface cards for desktops.
The transition to gigabit speeds in LANs gives rise to two related questions. First, is a fiber to the desk network cost-competitive? Second, can copper meet the challenge of higher speeds in the horizontal segment of the network? In my opinion, the answer to the first question is “yes.” The answer to the second question is “probably not.”
THE BOTTOM LINE
Long-term costs must be taken into consideration when calculating the cost effectiveness of fiber versus copper to the desktop. In the short term, copper may look attractive because of its perceived initial low cost. But when lifetime network costs are taken into account — installation, testing, maintenance and upgrades — the total cost gap between fiber and copper closes quickly.
Installation costs for fiber cables are close to those for Category 5 and 5e, and are less than those for proposed Category 6. This is due in large part to fiber’s strength and small size. The maximum pull strength for fiber cables is 200 pounds, versus 25 pounds for copper cables. In addition, fiber cables are typically smaller than copper. These characteristics make fiber easier and faster to install, and cut down on labour costs.
Still, one of the biggest obstacles to full-scale deployment of FTTD network cabling designs is the cost of electronics and connectors. Connectors alone have comprised 35 per cent of optical fiber cabling costs. Also, the bulk of conventional fiber connectors, which are twice the size of copper connectors, made comparable port densities impossible. Reducing the size of fiber connectors to enable increased port density and reducing their costs are keys to enabling the growth of FTTD.
CONNECTORS LOWER PRICES
A new family of fiber connectors, known as small form factor (SFF) connectors, is currently breaking down barriers and lowering costs in fiber networks. Compared to traditional fiber connectors, SFF connectors are smaller, less expensive and easier to install. And, as all SFF connectors are the same size as standard copper connectors, port density is doubled.
Doubling ports does not double the cost of electronics — most of the cost is in the required components, regardless of the number of ports. Therefore, greater port density on hubs and switches lowers the cost per port, making FTTD solutions more competitive with copper.
Fiber systems with SFF connectors cost much less than traditional fiber designs and approximately 20 per cent more than Category 5 copper solutions. The bigger the installation, the greater the savings and the narrower the gap with copper.
For example, one manufacturer prices a 200-station fiber installation using SFF connectors at US$40,550 compared with US$48,200 using standard fiber connectors and US$28,350 with Category 5. However, when the installation increases to 1,000 stations, the price tag for a small connector system is US$334,500, compared with US$410,000 with standard fiber connectors and US$332,500 with Category 5. In the larger installation, the cost difference between copper and FTTD is negligible.
The cost of fiber network hardware is coming down as well. For example, Fast Ethernet fiber network interface cards (NICs) can be purchased for about US$120 (Gemflex Network list price), less than half the price of two years ago. Media converter prices also have dropped by half. Moreover, with a centralized optical fiber cabling design, fiber is less expensive than copper to install and maintain.
Testing a fiber network is simpler and faster than testing copper. A minimum of four tests are required for Category 5 copper: wire map, length, attenuation and near-end crosstalk (NEXT). For GbE links, one must add tests for delay skew, power sum NEXT, power sum ACR, far-end crosstalk and return loss. With fiber, only one test (attenuation) is required.
The advantages of optical fiber are well known by now. With its very high bandwidth and low attenuation, optical fiber accommodates all current and envisioned network protocols, and does so without expensive and time-consuming re-cabling. The bandwidth of standard multimode fiber using 850 nm light-emitting diodes (LEDs) — 160 MHzkm for 62.5 m fiber, 500 MHzkm for 50 m — is designed to support today’s LAN applications. Now, as fiber manufacturers continue to advance multimode fiber system performance, low-cost lasers and laser-optimized multimode fibers are providing the means to transmit data at multi-gigabit speeds over distances required by tomorrow’s LANs.
Fiber provides great flexibility for network designers who take advantage of its longer link distances and implement an all-optical fiber centralized design. With all data electronics housed in a single location, optical fiber cables provide direct connections to every workstation outlet in the network in one of three ways:
1) “Pull through” or “home run” cabling, which involves passing cables through intermediate closets to the desktop without the use of patch panels.
2) “Splice”, which involves splicing to jumper cables in the intermediate closet, then linking to the desktop.
3) “Passive patch panel”, which uses jumper and intra-building cables in the intermediate closet.
COMPARING NETWORK DESIGNS
In the conventional, de-centralized premises data network design, backbone cables travel from a main cross-connect (MC) — or in an interbuilding network, an intermediate cross-connect — to one or more horizontal cross-connects (HC) within telecommunications closets (TC) on each floor of a building. The HC typically includes active electronics equipment — hub, concentrator or switch. Individual outlets for each user are located within 100 metres of the TC and are connected to the HC using a single cable per user in a physical star configuration (see Figure 1).
In this conventional design, most inter- and intra-building backbone cable is optical fiber; the horizontal segment of the network is typically comprised of unshielded twisted-pair (UTP) copper cable. The transmission distance limitations inherent in copper make the distributed design a necessity, because using copper in the horizontal requires that data electronics be located no more than 100 metres from workstations.
Designed to provide maximum flexibility in the deployment of distributed electronics, this infrastructure design has some drawbacks. Decentralization increases complexity and presents multiple potential points of failure. TCs take up valuable real estate, and, because of the active electronics, require power, air conditioning and grounding. The use of copper cable in the conventional design also places bandwidth limitations on the network. Due to its inherent electrical properties, copper is also vulnerable to EMI, RFI, crosstalk and breaches in data security.
The optical fiber centralized design provides direct connections between hundreds of workstations and a single MC, by using pull-through cables or a splice or interconnect in the TC (see Figure 2).
With network electronics, analyzers, uninterruptable power sources (UPSs), cross-connects and servers consolidated in the MC, the centralized design is a vehicle for reducing the number of TCs. The long cabling runs typical of these designs (often exceeding 100 metres) are perfect for fiber but impractical for copper.
rect connections between network hardware and desktops simplify maintenance and trouble-shooting. Speed upgrades are quick, easy and economical — merely a change of ports at the hub and NICs at the desktop computer. The passive patch panels require little real estate. And, unlike intermediate distribution frames containing active electronics, passive patch panels require no power, air-conditioning or grounding.
Finally, installers of optical fiber high-speed networks benefit from established standards. Gigabit Ethernet (IEEE 802.3z) was standardized in June 1998, and efforts are well along to complete a 10 GbE standard (IEEE 802.3ae). The standard for GbE over copper took a year longer to develop than the standard for fiber, a consequence of the difficulty of using copper at high speeds.
Previously, Ethernet transmitted over only two strands of copper, one on transmit, the other on receive. The remaining copper strands were used as grounds to promote isolation. With GbE over copper, all of this changes. None of the strands are used as a ground; instead, the strands are used both for transmit and receive. The electronics in the NICs and switches are supposed to manage the transition from transmit to receive and back again.
PROBLEMS ON THE HORIZON
There are various problems with Category 5 cable in GbE systems. The installed base of Category 5 may not be up to the demands of GbE’s new full duplex transmission on four wire pairs, particularly due to poor installation practices. But other concerns have been raised over confusing labelling, non-compliant connecting hardware, poor interoperability among products and inaccurate test equipment.
One concern is not new: problems with data transmission over copper cable are exacerbated at higher speeds. Most critically, EMI/RFI become worse as bit rates increase and copper conductors become better antennas, resulting in network disruptions. Now, GbE is giving rise to new concerns for Category 5 cabling. Full duplex transmission on all four copper pairs produces a critical new parameter, far-end crosstalk (FEXT). Old parameters — NEXT, return loss, power sum ACR and delay skew — all become crucial with GbE. Installed Category 5 cable may not comply with these new or more critical requirements.
In addition, concerns have been raised about the quality of Category 5 cable manufactured since 1994, when demand for Category 5 surged. Increased demand led to shortages of FEP, a critical insulation material, which was subsequently used in smaller quantities and mixed with other materials of varying electrical performance characteristics. Those cables made with less FEP are likely to cause problems.
QUESTIONS AND CONCERNS
Uncertainty about the performance of the installed base of Category 5 cable can be traced to installation practices that were in place before the introduction of high-speed networks. For example, 1000BASE-T requires that copper cables not be untwisted more than 13 millimetres prior to inserting wires into punchdown blocks. But cable installers have routinely stripped and unravelled copper pairs beyond the regulation 13 millimetres in order to punch them down.
In addition, installers have not always adhered to the restrictions regarding Category 5 maximum bend radius (no less than four times the cable’s diameter) or the over-cinching of cable bundles using tie wraps. These practices will have critical consequences in GbE systems running on installed Category 5 cabling. Cable installers can not be expected to follow a standard that did not exist at the time. But this means that Category 5 cabling installed prior to the standards may not perform as anticipated.
Category 5 cabling systems installed prior to 1995 and the completion of ANSI/TIA/EIA 568-A, may contain connecting hardware that is not compliant with that standard and may present further difficulties for a GbE network. Connectors and wall-plate terminations that do not meet top-grade Category 5 cable performance standards could degrade system performance with higher speeds. These deficiencies may lead to increased numbers of bit errors and garbled information, necessitating the retransmission of data.
Problems with hardware may be worsened by poor interoperability among products from different vendors. For a high-speed system to perform reliably at capacity, all connectors, cross-connects, patch panels and outlets must be capable of performing at the GbE speeds. The speeds specified with the Category 5 cable are only as reliable as the sum of the components used and the quality of the installation.
These and other questions about the ability of Category to meet the demands of GbE systems remain to be answered. Although many organizations continue to wonder if their Category 5 cable will perform reliably at gigabit speeds, many hold on to copper technology because of the perceived high initial investment to upgrade to fiber. Yet, taking into account the total cost of a network over a decade, including re-cabling with enhanced copper cables as data rates soar, installing optical fiber throughout the network makes sense.
Network owners can now take advantage of the falling costs, increased reliability and ease of installation and maintenance that an all fiber network offers.CS
Preston Buck is the Market Manager-Premises in the Telecommunications Products Division for Corning Inc., Corning, NY. In this position, he is responsible for anticipating and addressing the needs of the premises market. Before joining Corning in 1996, Mr. Buck was a U.S. Army officer who designed and developed data networks in military medical construction projects.