Jun 00 Getting Started
Volume Number: 16 (2000)
Issue Number: 6
Column Tag: Getting Started
Networks 201 - Part 2
by John C. Welch, Edited by Ilene Hoffman
Topologies, Signals, and Wires, Oh My!
In the first article of the series, we went through the OSI seven-layer network model, and discussed the purpose of each layer. We also covered the philosophy and reasons for the OSI model. Finally, we pointed out that there is a pseudo-layer 0 in the model, that deals with the areas between stations or computers. This layer 0 is the actual wiring or RF that connects the stations together, and this is what we will delve into this month.
As we talked about in the OSI article, Layer 0 is the media between stations in a network. Although Layer 1 is the Physical Layer, it actually only deals with the signals on the network, and ceases to function once you leave the transmitter, only picking up its function at the receiver. Layer 1 is blind to everything in between nodes on a network. This is actually good, because it gives us a lot of flexibility we might not otherwise have in media types. This is not to say that Layer 1 is completely independent of the media. If this were the case, then we could use Ethernet wiring and Token Ring network cards. Rather, as long as the connection interface between the media and the end node is correct, Layer 1 assumes that everything else is correct. So Layer 0's scope is all the wired and wireless media between node A and node B. In this article, we will stay on wired media. I covered 802.11 wireless networking in detail in the December 1999 MacTech, and recommend you read that issue for information on wireless media and topologies.
Logically, the first thing to discuss in a wired network is the wire. Wire is a bit of a misnomer, as when talking about copper network cabling, we are actually speaking of cable bundles. If we talk about fiber optic cable, then wire doesn't apply at all, beacuse in a fiber network, there is no wire. For ease of use, we'll use wire to mean any physical path for a network transmission to travel on.
Copper network cables are the most common, even though fiber optic cable has superior characteristics. Primarily, copper is cheaper than fiber on a per-foot basis. Considering the total length of even a small network can easily reach well over a mile when all is said and done, this is an important issue. Also, even with Gigabit networks being implemented, copper still works quite well. Copper is more simple to work with than fiber, and it's much easier for a network administrator to quickly build copper cable when needed, than it would be for fiber.
In spite of being somewhat harder to work with, and more expensive to buy, fiber has some clear advantages. It is capable of much higher speeds than copper, it is physically smaller, and it is immune to most electromagnetic situations that will cause great harm to copper networks. As an example, when I was an Administrator for a city in Florida, we had a building that was poorly grounded. Every time there was a lightning storm, (in Florida that means daily), we had to replace the 56K WAN modems. After a year of this, we talked the city council into budgeting for a fiber optic connection to the building. Once that was done, it was the most stable building in the city.
Since copper is the most prevalent wiring type, let's look at that first. There are three basic types of copper wiring for networks: Coaxial cable, shielded twisted pair or STP, and unshielded twisted pair or UTP. Coaxial cable, referred to simply as coax, is a round, fairly thick cable with two conductors laid along the same axis, (hence the name Co-Ax). The inner conductor is a single copper strand, with a layer of insulating material around it. This is covered by a third layer of braided conducting material, such as aluminum, and this layer is covered by a coating of PolyVinylChloride (PVC), or Teflon. A cross-section is shown in Diagram 1 below
Diagram 1. Coaxial Cable Cross-Section.
Coaxial cable has several advantages because of its construction. It is virtually immune to noise from outside sources such as fluorescent light fixtures and power lines. Coax also has excellent bandwidth capabilities. It has a ceiling of 2Gbps as long as the total cable length is less than 1km. These characteristics made coax a popular choice for network cable for many years. In fact, the original media specifications for Ethernet specifically required coax cable. Unfortunately, coax has a number of disadvantages which have led to its removal as a common network media. Coax is relatively fragile because of all its layers. It can be damaged or broken if it is bent on too small of a radius, or kinked. If it is subjected to even moderate crushing force, from say, a file cabinet, it easily breaks. Coax is also heavy, which made it difficult to use in a large network, due to the cost of supporting the wire bundles, whereas twisted pair wire can normally be laid on top of a suspended ceiling. Also, coax tends to be fairly thick, with diameters from 3/8 of an inch or more being common. This causes problems in dense installations, as available space for wiring runs and closets is quickly consumed. In addition, the complexities of manufacturing coax made its overall cost much higher per foot than twisted pair wiring. Finally, coax cable is not flexible in its implementation. Due to the need for impedance matching and proper termination of the lines, coax networks are very sensitive to breaks in the line. Since most coax networks use a daisy chain style implementation, any break in the line can easily bring down an entire physical segment of the network. This sensitivity also means that adding stations to an existing segment either requires bringing down part, or all of that segment, so that the new station and wire can be added in, or the new station must 'tap' into the line. Tapping into the line is where the connection for the node is clamped over the coax line, and a conducting spike in the connector punches into the conductors of the line, providing a connection point for the new node. When compared to the ease of plugging a node into a twisted pair network, this made continued use of coax hard to justify.
This is not to say that coax has no place in a modern network. In situations requiring high bandwidth, where fiber optic lines are not an option, coax is used extensively. The most familiar of these situations is for users of cable modems. Coax cable's high bandwidth and clean transmission characteristics make it ideal for cable modems. Network and television signals can be run over the same line, and then filtered out at the appropriate destination.
The next type of network media is twisted pair. This type of wiring is two wires, twisted around each other. Multiple sets of twisted pair wires are then combined in wire bundles. The number of sets depending on the network requirements, from one pair, (phone networks) to eight pair bundles, (Fast Ethernet). The reason for twisting the wires is simple: Two long pieces of copper wire next to each other tend to be a much better antenna than a carrier of network signals. By twisting the wires, this ability to pick up ambient electromagnetic radiation is destroyed, and the wires can then easily carry only the signals we wish.
There are two types of twisted pair wiring: Shielded and unshielded. Shielded Twisted Pair (STP) wiring combines the lower costs and greater flexibility of twisted pair wiring, with the better shielding of coax. STP takes a standard twisted pair cable, and wraps the wires in a layer of foil or braided metal as an extra layer of insulation. Although this gives STP superior protection from Radio Frequency Interference (RFI) and from things like light fixtures and power lines, it also adds greatly to the weight, thickness, and cost of the cables. IBM was one of the few companies to push STP, and as a result, you only see STP wiring in older IBM installations.
Unshielded Twisted Pair (UTP) is currently the most widely used wiring type in modern networks. It is lightweight, cheap to manufacture and install, and capable of speeds from 2Mbps to 1Gbps. UTP is often referred to by different categories, such as Cat 3, or Cat 5, but in practice Cat 5 is the UTP type that you see most often. The surprising thing about UTP categories is that there is really no other measure for assigning cable to a given category other than that cable's performance. Once the cable is built, it is tested at different speeds. If the cable is capable of reliable speeds greater than or equal to 100MHz, then it is a Cat 5 cable. If the cable's maximum speed is 20MHz, then it is a Cat 4 cable. (Other criteria, such as twists per inch have been proposed, but they have been superceded by performance.) Currently, the only other category used besides Cat 5 is Cat 3, in older installations. Cat 1 and Cat 2 were obsoleted in 1995, and Cat 4 offered such a small performance increase over Cat 3 that it was never really used.
Regardless of the category of the cable, or the LAN type used, there are certain constants to keep in mind for UTP. The most important of these is the actual layout of the twisted pair wires in the connections. Normally, where devices such as hubs and switches are used, the type of UTP used is known as a straight connect cable. That is, lead 1 on one end of the cable connects to lead 1 on the other end, lead 2 connects to lead 2 and so on. This is also called a DCE to DTE connection. The computer connecting to the hub or switch is the Data Terminal Equipment, (DTE) side of the connection, and the hub or switch is the Data Communications Equipment, (DCE) side of the connection. A straight cable works for this type of connection, because the wires on the devices the cable connects have different functions. An example of this is shown below in Diagram 2.
Diagram 2. DCE to DTE connection for 10BaseT UTP wiring.
As the diagram shows, the physical devices on either end are wired so that the transmit on one end connects to the receive on the other. This allows data to flow correctly. Although the diagram is for a 10Mbps Ethernet connection, other network types would look the same.
In situations where a DTE to DTE connection is needed, such as temporarily connecting two laptops together, or two hubs together, the straight type of cable won't work, so we use a crossover cable. This type of cable is named for the way the cables connect to different leads on different ends of the cable. The reason a straight cable won't work is because if two DTE devices are connected via a straight cable, you would have transmit leads trying to send to transmit leads, while the receive leads did nothing. The crossover cable resolves the problem by making sure that the transmit connects to the receive correctly, as shown in Diagram 3 below.
Diagram 3. DTE to DTE connection for 10BaseT UTP wiring.
Fiber Optic Cable
The final media/wire type is fiber optic cable. This type is different from copper-based media in both construction and signal transmission. Fiber optic cable, or just fiber, has a fairly simple structure. The outside is sheathed in plastic, the next layer, or cladding, is plastic, and the main conductor is high-purity glass or plastic. Fiber is extremely thin compared to UTP cabling, because the light waves that pass through it have a wavelength, or size measured in nanometers, or billionths of a meter. The cladding and sheathing are measured in microns, or millionths of an inch, so a bundle of fiber the same size as an average UTP cable has many more conductors than the UTP cable, and far more capacity. Usually, LAN grade fiber is known as 62.5 micron glass, (conductor) diameter fiber. You also will see it expressed as 62.5/125 micron glass, where the 62.5 is the conductor size, and the 125 is the diameter of the cladding. Fiber optic cables are used in pairs, each cable carrying light pulses in a single direction, either to or from a station on the cable. This is due to current limitations in using light as a transmission signal.
There are two basic types of fiber, multi-mode and single-mode fiber. Multi-mode is named due to an effect created by the light source. Multi-mode fiber uses an LED (Light Emitting Diode as its light source. When an LED burns out, it is called a DED, or Dark Emitting Diode, but these are less useful. LEDs are neither concentrated nor coherent light sources. That is, the light they emit is widely dispersed from the source, (not concentrated), and covers many different frequencies, (not coherent). This imposes some bandwidth and distance limitations on LED-driven fiber. Ideally, all the light from the LED would stay in the center of the conductor from the LED to the end of the cable. This is called axial travel, and is the desired way for the light pulses to travel within the cable. However, due to the lack of concentration of the LED's signal, this doesn't happen. Instead, parts of the light pulses reflect off of the sides of the conductor. Although these reflections contain the same signal as the part of the beam still traveling axially, because they are traveling a different path through the conductor, they are now known as modes. Also, due to this new direction of travel, the modes take longer to reach the end of the cable than the axial beam. This can cause problems on the receiving end, because instead of one single signal being received, the axial signal and all of the reflections are received at different times, which could appear to the receiver as different data than was sent.
While all this bouncing around is taking place, remember that each time a mode passes through the axial signal, it can cause new reflections to take place, creating more modes, and potentially more confusion. The upside to multi-mode cable is that it, and the equipment it connects are substantially cheaper than single mode cable. In most situations, the individual cable lengths are only a few meters, so attenuation caused by the modes is not an issue. Finally, multi-mode cable is thicker than single mode, so it is easier to work with. (If the interaction of the axial and reflected signals/modes seems confusing, this is due to the nature of light. Light can act as though it is particles or waves, depending on the context in which it is viewed. To thoroughly explain this would delve far deeper into advanced quantum physics than we have space or patience.)
The second type of fiber is single-mode fiber. This uses a true LASER as its light source, usually an Injection Laser Diode (ILD). ILDs have two advantages over LEDs: Concentration and Coherence. A laser tends to be a tightly focused beam of light, that has little dispersal over the distances used in a LAN. (Indeed, even at much greater distances. It is possible to send a laser from the Earth to the Moon, and have the dispersal diameter at the Moon be only about a foot.) This helps avoid the modes, and problems caused by modes that are normal for LEDs. Since LASERs, unlike LEDs, only emit a single frequency of light, they are said to be coherent. This allows for greater range, and also for the fiber manufacturers to better match the ILD wavelength with the fiber, for smoother data transfers. The end result of this coherence and concentration is that even over many hundreds of meters, the beam does not spread enough to touch the sides of the fiber conductor. The beam is able to stay in the center of the conductor, or have near-perfect axial travel, for the entire length of the cable. This allows all the information in the beam to arrive all at once, or in a single mode, hence the name. Single mode fiber conductors are usually between five and ten microns in diameter, with 125 micron cladding. Although single mode has greater range and data rates, it is also more expensive to manufacture, and install. Single mode is most often used in situations where long distance is a necessity, such as telephone network backbones. Currently the maximum distance for single mode is around 30Km, although with the use of repeaters, that could be extended to 100Km.
Since fiber optic cable requires light at a very specific frequency, or set of frequencies to work, it is worth our while to take a look at some of the characteristics of that light. As it turns out, there are three specific frequency bands, about 25,000 to 30,000 GHz wide, that work the best for fiber. The reason for this is evident when you graph attenuation against wavelength, as shown in Diagram 4.
Diagram 4. Attenuation vs. Wavelength for LASERs in fiber optic use.
As you can see, the two largest wavelength bands, at 1.30µ and 1.55µ respectively have the lowest attenuation across their width. If you go much larger than 1.55µ, the attenuation rises fast enough to cancel out any advantages gained by using a larger bandwidth. The reason for using the smallest wavelength band, the 0.85µ band is a material one. At 0.85µ, the lasers and the electronics using the lasers can all be made out of the same material, gallium arsenide, which makes manufacturing easier. The savings in manufacturing balances out the attenuation caused by the higher frequency. Remember, wavelength and frequency are inversely proportional. The bigger the wavelength, the smaller the frequency. So now that we have covered the wiring types, the next step is to deal with how they are laid out, or topology.
The topology of a LAN refers to the layout of the wires and devices on that LAN. There are two types of topologies, physical and logical. The physical topology is the actual physical location of the media and stations on a LAN. This includes wiring closets, cable guides, wiring runs, etc. The logical topology is what we refer to when we talk about rings, stars, or,, backbones. Most often, the physical topology is only used in a general sense, as specifically describing the physical topology of a large LAN is unwieldy at best. In this article, we will deal with generic physical topology types.
Topologies fall under four basic types: Bus, Ring, Star, and Switched. The first type, Bus, is the simplest of the four. It is essentially a long main cable, or backbone, that all stations connect to directly. Each end of the cable has terminators. The terminators are needed so that any signal reaching them ends there, and does not reflect back onto the line. This is also known as Voltage Standing Wave Ratio, or VSWR. (There are two reasons for avoiding VSWR. The first is that the reflections can induce errors in the data flow. The second is that the reflection can damage equipment. In higher power uses, such as RADAR, VSWR can actually melt down equipment.) Stations on the Bus transmit frames in all directions. If the frame reaches a station that it is meant for, that station pulls the frame from the line. If not, the frame continues until it reaches the terminator. Although simple to set up, and inexpensive, the Bus topology has been left behind for a number of reasons. It's inflexible, because adding a station means physically modifying the backbone line via taps. This limits the number of stations on a line because you can only insert so many taps before the data degrades. This type of physical connection also makes it hard to connect backbones together. The backbone itself is a single point of failure, because if it is cut or damaged, all stations on that backbone are unable to use the network. The bandwidth is shared among all stations, so if you have ten stations on a 10Mbps backbone, each station ends up with a 1Mbps connection. Finally, the backbone cable itself has to be fairly thick, and unwieldy, so as to be easily tapped.
The second topology type is the Ring. At its most simple implementation, it is similar to a Bus, in that all stations directly connect to the ring. In that configuration, it also shares most of the Bus's disadvantages too. However, in more modern uses, such as Token Ring, and FDDI, certain modifications are made to the physical topology, but the logical topology is still a ring. Using Token Ring for an example, the Ring itself is a set of two conductors, either copper or fiber optic. The conductors attach to boxes called Media Access Units or MAUs. The MAUs function like hubs in an Ethernet network. The Ring conductors connect to ports labeled Ring In and Ring Out, indicating the direction of data flow. Each MAU has a number of ports that individual stations connect to, such as computers or printers. Each Ring can have multiple MAUs, just as an Ethernet network can have multiple hubs. By using MAUs, the Ring gains a great deal of flexibility that it would otherwise not have. Since stations can be attached or removed to the MAUs as needed, reconfiguring segments of the network is simplified. In this configuration, the Ring resembles a Star topology, although electronically, it is still a Ring. Also, MAUs can come with two Ring In and Ring Out ports each, allowing for redundancy. The Ring topology is not the most common, mostly because it is only needed by a small number of network types, the previously mentioned Token Ring and FDDI among them. Like the Bus, bandwidth is shared among all stations on the Ring.
The third topology is the Star. This is the most common, and most familiar LAN topology in use. In a Star setup, all devices connect to a central hub. Each device accesses the LAN independently of the others. That is, unlike a Ring or Bus, on a Star, there is no backbone that carries all the data to all the stations. The hub functions as a pseudo–backbone, but without the electronic requirements of a Ring or Bus. In a Star, hubs can be interconnected, to allow for greater flexibility of configuration and load balancing. The number of stations on a given Star is limited by the size of the hub, and available bandwidth. Similar to the Bus and Ring, bandwidth on a Star is shared by all stations.
The final topology, Switched, is not so much a separate topology as a different implementation of existing topologies. As we noted in the three previous topologies, they have one common problem: Shared bandwidth. Regardless of how much bandwidth you start out with, it is split among all stations on that topology, resulting in much slower performance than desired. One of the solutions to this is, the switch. The switch allows each station to have full access to the available bandwidth, at whatever speed the switch provides to that connection. As an example, in a Star network, if the hub is capable of a maximum speed of 100Mbs, and has 24 ports that are being used constantly, then each port only has access to 4.17Mbs of bandwidth. Without a switch, the only solution is to buy more hubs, and place less load on each hub. However, with a switch, this changes. If you have a 24–port 10Mbps switch, and the switch connects to a 100Mbps network, then each of the 24 devices on the switch has access to a full 10Mbps of bandwidth. Even if the switch is on a 10Mbps network, each port still gets a full 10Mbps connection. This is because as each station on the switch transmits or receives, for the instant that the station is handling packets, it is the only station accessing bandwidth on the switch. In other words, the switch creates a temporary connection with only two devices on the network. This gives full access to bandwidth, and improves speed and reliability by reducing collisions. Any topology can be converted into a Switched version for improved performance. Another advantage of the Switched topology is it creates virtual networks, and tracks these networks with internal tables based on things like MAC addresses, The Switched topology allows for greater configuration and control of traffic, through such technologies as VLANS, which we will look at in more detail in a later article. The disadvantage to a Switched topology is primarily expense, as switches are more expensive than hubs or MAUs.
Although we dealt with quite a few wiring types and topologies, we have covered the scope of Layer 0. Normally, you are not going to encounter STP wire, or Bus topologies in your daily work. Coax was moving out of favor, but with the rise of cable modems, and broadband networks, you are likely to have to deal with it in the future. I would also recommend doing some independent study on fiber optic networks, as they are becoming more standard as network speed and reliability needs increase. As far as topologies are concerned, the Bus is rarely encountered in modern networks, but that could change as well. Stars, followed by Rings are the current main topology types, with Switched versions of each being extremely common. (Actually, there is a case for the Switch being the most popular topology type, but that depends on who you ask, and what vendor they use.) In any case, switching networks are rising rapidly due to their bandwidth efficiency and flexibility of configurations.
Bibliography and References
- Tannenbaum, Andrew S. Computer Networks. Third Edition Prentice Hall, 1996
- Sportack, Mark. Networking Essentials Unleashed. SAMS Publishing, 1998
- Stallings, William. Local & Metropolitan Area Networks. Fifth Edition Prentice Hall, 1997
John Welch <firstname.lastname@example.org> is the Mac and PC Administrator for AER Inc., a weather and atmospheric science company in Cambridge, Mass. He has over fifteen years of experience at making computers work. His specialties are figuring out ways to make the Mac do what nobody thinks it can, and showing that the Mac is the superior administrative platform.