Network Design And The Choice Of Physical Media
Introduction
Not so many years ago the only way to get information from one computer to another was using manual methods. An individual had to transfer data contained in some physical media from one computer to another, or the information had to be manually reentered by hand. Today this information is transferred at a high rate of speed, using sophisticated networking protocols, and with an almost immeasurably low error rate.
Networking today is something we almost take for granted; yet not so long ago it seemed remarkable. Today we print to a printer located in another room. We send an email to a colleague in another building. At home we listen to Internet radio while sending instant messages to our friends. Later we browse sites created by people located halfway around the world. Very cool.
Arguments rage over the best Internet browsers. People compare notes on the best communications software. Magazines are devoted to one specific network operating system. The books and certification programs devoted to networking protocols are too numerous to count. But the physical infrastructure is invisible; the network connection is just one more wire protruding from the back of the computer.
But the network didn't just happen. It wasn't willed into being. It was designed, it was built and it physically exists. Just as a wise man builds his house upon a rock, the wise network designer pays careful attention to the choice of the networking media.
Yet wires aren't sexy. Wires aren't cool. The wires are everywhere, and since we can see them, touch them, and perhaps trip over them, we think we understand them. The sexy part about networking is the stuff that you can't see and that the average person doesn't understand: the applications, protocols and network operating systems that move all those ones and zeroes around at the speed of light. Knowledge of the sexy bits gets you recognition from your peers and a nice fat paycheck to boot; knowledge of the mundane is despised.
This is unfortunate, because the failure to understand the characteristics of the physical layer may result in an improper installation and a poorly performing network. Some physical media are inherently more secure than others. Some transmit longer distances or at higher speeds. Some are less expensive. The type of physical media chosen affects the size of the data packets, and can have dramatic impact upon the way certain protocols, network operating systems, and communications applications should be configured and operated.
Because the physical layer is ignored and the skills required to build and maintain it are disregarded, very little systematic information is available. It is easy to find information comparing different network operating systems, including their respective costs, reliability, and supportability. It is much more difficult to make the same side-by-side comparisons of different physical media.
Suppose you need to install a new network. Should it be fiber? Should it be cable? Should you go wireless? What kind of fiber? What kind of cable? Which implementation of wireless? The network designer has a multitude of choices to make with very little solid, comparative information to help make that choice. Wireless is a relatively new media for a LAN. The ink is scarcely dry on the standards, and it has not been proven in a corporate environment. We will focus primarily on conducted media---wires and cable---and assist the network designer to make informed, intelligent decisions.
Getting Physical
There are two basic Categories of physical media: conducted media and radiated media. In conducted media the signal is conducted by or carried over the circuit by the conductor. Radiated media, instead of using conductors, transmit the signal through space between the transmitter and receiver. These types of systems are sometimes called spacewave or free-space systems (Horak, 2002).
Conducted media are generally of two types: wires transmitting electrical signals, and fiber optic cables transmitting signals using light waves. In general, wires are used for short-range, lower speed applications, while fiber optics are used for long-range, high speed communications. These general distinctions don't hold up to close examination, as some high end wiring media are as fast as low end fiber optic cabling, (although the maximum lengths of high speed wiring media is one third that of low end fiber optic cabling.)
Everything else being equal, bandwidth and maximum transmission distance are inversely proportional.[1] Bandwidth is also more expensive. The network design must ensure the physical media can support the necessary bandwidth over the required distance, be maintainable, and be within budget. The network must also be secure, and the need for security may preclude certain media from consideration, (at least within the LAN itself.)
Some professionals seem to believe it is important to choose a category of physical media before factoring in bandwidth and distance; at the very least they recommend selecting from the various implementations of a media category based on bandwidth and distance (Wilson, D. J.). This would seem to indicate that the costs of installation and maintenance should be major considerations in any network installation.
But costs of installation and maintenance, although important, do not alone dictate the type of physical media used. Any choice of physical media must begin with the "data rate…expected of the LAN" (Trulove, p. 21). The choice of standards based structured wiring meeting the required data rate will then be dictated by considerations of the costs of installation and maintainability.
Cost Factors:
Given that the return on investment is likely the primary consideration for an organization, the cost of the network is a major consideration in its design. Many different factors determine the network costs: materials; installation; tools & test equipment; and training. One important cost factor is whether the technology is widely available; one good measure of this is when the technology begins showing up in the larger retail computer stores. When the product volume is such that the average consumer can purchase it, the technology is well established and the cost is minimal.
Any decent computer store will carry two basic kinds of networking media: 10/100BaseT, (generally Cat 5 UTP cables, connectors, NIC cards, and crimpers,) and wireless media, (WiFi, more formally known as 802.11b.) It is possible to find some odd home networking gear like power line or phone line networking, but that appears to be fading when compared to wireless.
The cost of the physical media themselves becomes a consideration, especially as the distances and numbers of connections increase. While it is commonly thought that fiber optic cable is more expensive than UTP, a comparison of today's prices shows this not to be the case. The cost of Cat 5 cable is $0.09 per foot. Cat 5e is $0.075 cents per foot. Single mode fiber optic cable runs $0.08 per foot. The price of Cat 5, Cat 5e and single mode fiber, then, are roughly comparable. UTP becomes more expensive than single mode fiber only if you want to install the latest Cat 6 cable. UTP, (with the exception of Cat 6,) and fiber optic cable are generally less expensive than either thinnet or thicknet coax.
Table 1: Wiring Comparison
|
Type of Wiring |
Price
per foot |
Comments |
|
CAT3 UTP |
$0.03 |
Broadband
signaling |
|
CAT5 UTP |
$0.09 |
Broadband
signaling |
|
CAT5e UTP |
$0.075 |
Broadband
signaling |
|
CAT5e enhanced
UTP |
$0.12 |
Broadband
signaling; Tested to 400Mhz |
|
CAT6 UTP |
$0.18 |
Broadband
signaling; Tested to 550Mhz |
|
Coax |
$0.16 |
Baseband
signaling |
|
Coax Thinnet |
$0.13 |
Baseband signaling |
|
Fiber, multimode |
$0.22 $0.13 |
Dual fiber
zipcord Distribution
cable, 12 pr |
|
Fiber, single
mode |
$0.10 $0.08 |
Dual fiber
zipcord Distribution
cable, 12 pr |
NOTE: Price comparisons made at ControlCable.com, (http://www.controlcable.com), Datacom Link, (http://www.datacomlink.net), and FOnetworks.com, (http://www.fonetworks.com). Prices per cable/per foot for fiber optic cabling drop significantly when using multi-fiber distribution cable.
As for the NICs themselves, the cost of fiber optic NICs is a minimum of five times that of UTP NICs. A standard UTP NIC can cost around $25, while a NIC with both fiber and UTP connectors generally costs over $125.[2] Oddly, NICs with only fiber connections were even more expensive than dual use NICs.
The cost of tools and test equipment should also be factored into the decision. A basic tool kit for installing and performing basic testing of CAT 5 connections costs around $150; a tool kit for installing fiber optic connectors costs over $800.[3] The cost of a reasonable fiber optic test kit, including adapter cables and software, can range from nearly $2,000 to just under $3,000 (Lenny). Depending on the type of fiber and the level of analysis to be performed, the prices can continue to climb. Frankly, there are too many variables involved with cable to nail the costs down much further. On the other hand, the full-featured Fluke OptiView Integrated Network Analyzer can cost upwards of $16,000 (CNET). While this is debatable, it would appear the average cost of test ordinary network test equipment is somewhat less for UTP than for fiber optic cable.
The cost of the physical medium is clearly a major issue. If we had to install 10,000 feet of cable, the difference in cost between the low and high cost solutions, (Cat 3 and multimode fiber,) is $1,900; the difference between Cat 5e and Cat 6 is $1,050. This cost difference doesn't factor in the differences in connectors, tools, and test equipment between the different standards.
Clearly if cost alone was the object, installing CAT 3 UTP would come out on top. Obviously cost alone does not dictate network design, because CAT 3 UTP networks are hard to find. But purchasing on bandwidth alone is quite expensive. "One of the drawbacks of living on the bleeding edge is that you always pay too much for [it] and end up stuck with it once better stuff becomes available" (Thompson). Since we need to strike a balance, let us examine other important factors.
Maintenance Factors:
When determining the requirements for a new LAN, maintainability should always be kept in mind. This is done partly by adherence to standards, but maintainability should account for the future supportability of the physical cabling standard. CAT 3 cable is still available, but is getting harder to find. We can expect to see CAT 3 cable disappear the way CAT 4 did some years ago. But CAT 5 cabling is now more expensive than its enhanced sibling, CAT 5e, and we can expect CAT 5 to gradually disappear as well. Although CAT 6 cabling is a more capable medium than CAT 5 or CAT 5e, it is also more difficult to work with and requires more expensive equipment. Since the physical medium may be in place for twenty years, some consideration should be given to ensuring the pieces to support the infrastructure will still be available. Using the latest technology should help.
Some thought should be taken to how strong the basic physical media and connectors are. Of course fiber optic cable, being made of glass, is breakable. It is not as fragile as it might seem, but some care does have to be taken with it. At the other extreme we find coaxial cable. Coax cable itself is extremely strong and rugged; however, the coax connectors are actually quite weak and the most likely source of failure (Thompson). UTP cable is strong like coax cable, but its connectors are much more rugged than the BNC connectors used with coax. From the standpoint of strength and reliability, UTP is the best choice.
The physical network should be invisible. That means the physical installation should just work. The cables and connectors should not be a source of problems, nor should the way the cables were installed make it difficult to reconfigure the network. Fortunately standards have been established that, when followed, help ensure that the physical installation is rarely, if ever, the source of problems. These standards constitute an approach called structured wiring.
Structured wiring helps a designer decide how best to break up the network so no LAN topology violates the maximum allowable lengths (Trulove, p. 10). These individual units are combined into larger and larger interconnected network structures, ultimately becoming the total system itself. Standards exist for specific wiring technologies, such as 10BaseT, but the more generic standards, describing how cables themselves should be installed, are equally important.
Table 2
|
Unstructured |
Structured |
|
Temporary |
Permanent |
|
Over
the wall |
Behind
the wall |
|
Wire
as needed |
Wire
all work areas |
|
No
wall outlets |
Wall
outlets |
|
Free
form |
Defines
distances and topology |
|
Unplanned |
Planned |
|
Not
flexible (no cross-connecting) |
Flexible
(cross-connecting) |
|
Not
labeled or documented |
Labeled
and documented |
NOTE: Information retrieved from
http://www.rad.com/networks/1994/transmis/wiring.htm
A wiring standard provides
benefits to all aspects of the industry:
· To users, by providing an application-independent cabling standard and an open market for cabling components.
· To manufacturers, by providing a cabling system which accommodates current products and provides a template for future product design.
· To architects and the building industry, by providing the means to cable a building before the application requirements are known.
· To installers, by presenting guidelines for planning, installing and maintaining cabling systems.
In addition, an international standard avoids national trade barriers which is an important goal of the Commission of the European Communities (Building Wiring Standards).
The LAN designer, taking advantage of structured wiring systems, simply has to know the required data rate. "The data rate will define the category of cable…used" (Trulove, p. 21). The standards define the "horizontal wiring structure, the backbone wiring structure, and a series of cross-connect structures that include the main cross-connect, the intermediate cross-connect, and the horizontal cross-connect" (Trulove, pp. 22, 23). Everything involved---distances, cable grades, connector types and installation practices---is defined in the structured wiring standards (Trulove, p. 23). Following approved standards ensures the installation will function correctly when installed and will remain stable over time.
So maintainability has to do with the availability of components, the strength of those components, and the proper choice and installation of those components. A physical LAN, installed as a structured wiring system, is inherently maintainable. A physical LAN, when installed as a structured wiring system, is also inherently unlikely to require maintenance.
Figure 1: Hierarchy of Structured Wiring (Trulove, p. 24)
Security Factors:
A wire can be considered as an antenna. Electrical and magnetic fields, when crossing a wire, cause electricity to flow along that wire. Shielding can help prevent these fields from reaching the signal wire by ensuring the fields affect the shield only, and by ensuring the resulting electricity is shunted to ground. Wire twists have the same effect by forcing the created electrical signals to flow in opposite directions along the same wire, thereby canceling themselves out and leaving the original signal alone.
But an electrical signal flowing along a wire produces electrical and magnetic fields. These fields can be detected from quite a long distance away. If the electrical signal contains some sort of embedded intelligence, that intelligence becomes part of the electrical and magnetic fields, and can be detected and deciphered. Fortunately, the same effects that prevent electrical and magnetic fields from affecting the electrical signal also help prevent the electrical intelligence from being emanated. Help prevent their emanation, not prevent it entirely.
For some network installations the risk of compromising emanations is too high. For these installations fiber optic backbones and distributions are the natural solution. Fiber optic systems carry intelligence on beams of light. This light stays within the cable and has no compromising emanations.
But fiber optic cable has another important security feature: it is hard to tap. The installation of a tap into a wired network is fairly straightforward. Such a tap would be easy to detect with a time domain reflectometer, (TDR,) but using a TDR on a network is ordinarily only done to detect problems. It is not ordinary procedure to use a TDR to detect a wiretap. To do so one has to have an accurate map of the network, to know the exact distances of each leg and the devices installed.
Fiber optic taps, while not unheard of, are much
harder to perform. (Rumors exist of
Security costs. As we have discussed previously, the costs of the fiber optic cables themselves have dropped. Still, various other fiber optic networking components and test equipment are quite expensive. Fiber optic cable costs in maintainability---the cable and its connections are more fragile than UTP, so installing fiber optic cable is likely to result in more trouble calls for bad cables. The security needs must be weighed against the costs. If the benefits outweigh the costs, then the use of fiber optic media to the desktop is recommended.
Distance Factors:
As mentioned before, the data rate required dictates the type of cabling to be installed. The maximum distances of each leg of the network are well known and covered by the wiring standards. Still, there may be some installations where the distances involved dictate a change of physical media. It is important to know the maximum distances each technology is capable of at the desired bandwidth. We previously mentioned that doubling the distance of a leg cuts the maximum bandwidth in half. Network problems can therefore be caused by not strictly adhering to distance limitations.
The fiber bandwidth-distance product determines the usual distance of a link by "dividing the [bandwidth] by the distance of the link" (Trulove, p. 75). The higher the bandwidth required, the shorter the distance of each leg. "The bandwidth-distance product of 500 Mhz-km yields an estimated usable bandwidth of 500 Mhz at 1 km, 1000 MHz at 500 m, and 2000 Mhz at 250 m." (Trulove, p. 75).
The bandwidth of fiber optic cable comes down to the choice of a standard; the standard then dictates the type of transmission and reception devices. At the bandwidths required in an ordinary LAN installation, a single fiber leg can be quite long. Few Gigabit LANs, for example, would require a single leg to be longer that 500 meters. If such a leg were required, then the choice would be to use a different, more effective (and expensive) encoding standard on that leg.
While the bandwidth-distance product has a dramatic impact on available bandwidth, once again the use of standards eliminates any guesswork. If the required data rate is known, the standards themselves, (FDDI, Gigabit Ethernet,) will dictate the maximum distances.
Table 3
|
Type of Wiring |
Maximum Length |
Comments |
|
CAT3 UTP |
290 ft |
Broadband
signaling |
|
CAT5 UTP |
328 ft |
Broadband
signaling |
|
CAT5e UTP |
328 ft |
Broadband
signaling |
|
CAT5e enhanced
UTP |
328 ft |
Broadband
signaling; Tested to 400Mhz |
|
CAT6 UTP |
328 ft |
Broadband
signaling; Tested to 550Mhz |
|
Coax |
1640 ft |
Baseband
signaling |
|
Coax Thinnet |
607 ft |
Baseband
signaling |
|
Fiber, multimode |
? |
Length depends on
the standard, not the cable |
|
Fiber, single
mode |
? |
Length depends on
the standard, not the cable |
NOTE: Distances taken from LAN Wiring by James Trulove and from Cabletesting.com, a service of Fluke Networks, Inc.
Bandwidth Factors:
The bandwidth of a physical media is expressed not
in bits per second, but in hertz, or cycles per second. Thus bandwidth, expressed in hertz, is
different from the data rate, expressed in bits per
second. The way data bits are encoded
onto the sinusoidal waveform results in a high number of bits per cycle;
therefore, the data rate is higher than the bandwidth (Trulove, p. 139).
Table 4
|
Type of Wiring |
Operating
Frequency |
|
CAT3 UTP |
20 Mhz |
|
CAT5 UTP |
100 Mhz |
|
CAT5e UTP |
100 Mhz |
|
CAT5e enhanced
UTP |
100 Mhz |
|
CAT6 UTP |
200 Mhz |
|
Coax |
Same as CAT5 |
|
Coax Thinnet |
Same as CAT5 |
|
Fiber, multimode |
160 Mhz (minimum) |
|
Fiber, single
mode |
500 Mhz (minimum) |
NOTE: Bandwidth comparison taken from LAN Wiring
by James Trulove and from
Cabletesting.com, a service of Fluke Networks, Inc.
A
specific physical media is certified to a particular bandwidth. For example, Cat 5 cable is certified to 100
Mhz. Due to the type of encoding used by
the 100 Mbps Fast Ethernet standard, only 30 Mhz of the total available
bandwidth is used to deliver the 100 Mbps data rate. Using the ATM 155 standard allows a 155.52
Mbps data rate while using less than 30 Mhz of bandwidth. Using a different encoding standard and a
greater portion of the cable's maximum bandwidth allows for Gigabit Ethernet
over the Cat 5 and 5e standards. Cat 6
cable, certified at 200 Mhz, has potential for data bit rates well in excess of
today's Gigabit Ethernet.
It turns out that, unlike copper wiring, the fiber optic cables themselves have little effect on bandwidth; fiber optic bandwidth is more a function of the transmission and reception devices (Trulove, p. 74). This explains how newer multiplexing technologies and standards using different wavelengths of light are able to greatly expand the bandwidth of a single fiber.
Wireless
Three types of wireless installations exist; indoor point-to-multipoint, outdoor point-to-point, and outdoor point-to-multipoint systems. Indoor point-to-multipoint, due to its short range, is used most often for installations where wires are impractical. Outdoor point-to-point is used most often where wired connections are too expensive, like between two skyscrapers. Outdoor point-to-point is capable of ranges up to 20 km, and is used in place of wired high-speed connections. Outdoor point-to-multipoint systems are used to provide wireless broadband connections to households within 8 to 10 km of the central access point (Trulove, pp. 359 - 362).
Indoor point-to-multipoint systems include the increasingly popular WiFi standard, known more formally as the 802.11b standard. (This is the equipment that is beginning to be widely available in computer stores.) Other indoor point-to-point standards exist: HiperLAN, from the European Telecommunications Standards Instute, is a multimedia installation capable of 24 Mbps; and Shared Wireless Access Protocol (SWAP) from the Home RF Working Group (Trulove, pp 343 – 358).
Both outdoor point-to-point and point-to-multipoint systems are really not LAN technologies. Outdoor point-to-point systems are used to provide a communications link between two LANs. The links themselves may be made with radio waves or through some type of optical (laser) link. Outdoor point-to-multipoint systems are used to provide broadband Internet connections to end users who may or may not be sharing that connections through a local LAN. Both of these types of systems are outside our scope.
While we typically say the data rate required dictates the standard used, indoor point-to-point systems are generally chosen for reasons other than data rate. One possible reason for choosing wireless is if the building contains asbestos and running wires is unsafe (Trulove, p. 345). Another reason is convenience---say, providing connectivity in a conference room. Some business offer wireless access as a service; Starbucks has found a way to meter wireless access and is making money selling access to its customers (Starbucks).
The wireless standards were never meant to contain a comprehensive set of enterprise level security tools (Wireless). It is possible to lock down a wireless network and keep all but the most dedicated and knowledgeable intruders out of the network, but only at the risk of limiting the most useful features of the network itself---mobility. Finally, the indoor wireless standards are much too slow when compared to wired networks. 100 Mbps wired Ethernet networks are not uncommon today, while 802.11 wireless networks operate at either 1, 2 or 11 Mbps, (depending on range and interference)(Trulove, pp. 354, 362, 363). Wireless access is useful in limited situations, but is much too slow and insecure for widespread business use.
Conclusion
The physical media chosen for a network, when installed as part of a structured wiring system, is dependent on the required data rate. The required data rate is a function of the applications to be run and the number of users. Once the required data rate is known, this will suggest the standard required. The standard required will then determine the type of installation.
It turns out, with very few exceptions, that the type of physical media is predetermined by the standard. Thus the main difference in cost will be in the degree of security and upgradability required. Since a physical installation must last for many years, the network designer must determine whether---in the lifetime of the physical media---the data rate required will exceed the capability of the installation. This decision will determine whether the minimum required media will be installed or whether the media must be upgraded to allow for future required capabilities.
Of course any upgrade in capability will increase costs, and costs are always important. The network designer must perform a cost/benefit analysis of the costs of an upgrade now versus the costs of upgrading later. Any installation must present a business benefit. No rational business enterprise will spend money on a technical solution just because it is "cool." Ultimately the client determines how best to balance the benefits against their costs.