Novell's Networking Primer
The term "network topology" refers to the layout of a network. Due to the specific nature of computer network technology, networks must be arranged in a particular way in order to work properly. These arrangements are based on the network hardware's capabilities and the characteristics of the various modes of data transfer. Because of these factors, network topologies are further subdivided into two categories: physical topologies and logical topologies.
The physical topology of a LAN refers to the actual physical organization of the computers on the network and the subsequent guided transmission media connections. Physical topologies vary depending on cost and functionality. We will discuss the three most common physical topologies, including their advantages and disadvantages.
The simplest form of a physical bus topology consists of a trunk (main) cable with only two end points. When the trunk cable is installed, it is run from area to area and device to device—close enough to each device so that all devices can be connected to it with short drop cables and T-connectors. The principal advantage of this topology is cost: no hubs are required, and shorter lengths of cable can be used. It is also easy to expand. This simple "one wire, two ends" physical bus topology is illustrated in Figure 10.
Figure 10: Physical bus topology
A more complex form of the physical bus topology is the distributed bus. In the distributed bus, the trunk cable starts at what is called a "root" or "head end," and branches at various points along the way. Unlike the simple bus topology described above, this variation uses a trunk cable with more than two end points. Where the trunk cable branches, the division is made by means of a simple connector. This topology is susceptible to bottlenecking and single-point failure. The distributed bus topology is illustrated in Figure 11.
Figure 11: Distributed bus topology
The simplest form of the physical star topology consists of multiple cables—one for each network device—attached to a single, central connection device. 10Base-T Ethernet networks, for example, are based on a physical star topology: each network device is attached to a 10Base-T hub by means of twisted-pair cable.
In even a simple physical star topology, the actual layout of the transmission media need not form a recognizable star pattern; the only required physical characteristic is that each network device be connected by its own cable to the central connection point. Like the distributed bus topology, this topology is vulnerable to single-point failure and bottlenecking.
The simplest form of the physical star topology is illustrated in Figure 12.
Figure 12: Physical star topology
The distributed star topology, illustrated in Figure 13, is a more complex form of the physical star topology, with multiple central connection points connected to form a string of stars.
Figure 13: Distributed star topology
Physical Star-Wired Ring
In the star-wired ring physical topology, individual devices are connected to a central hub, just as they are in a star or distributed star network. However, within each hub the physical connections form a ring. Where multiple hubs are used, the ring in each hub is opened, leaving two ends. Each open end is connected to an open end of some other hub (each to a different hub), so that the entire network cable forms one physical ring. This physical topology, which is used in IBM's Token-Ring network, is illustrated in Figure 14.
Figure 14: Physical star-wired ring topology
In the star-wired ring physical topology, the hubs are "intelligent." If the physical ring is somehow broken, each hub is able to close the physical circuit at any point in its internal ring, so that the ring is restored. Refer to details shown in Figure 14, Hub A, to see how this works.
Currently, the star topology and its derivatives are preferred by most network designers and installers because these topologies make it simple to add network devices anywhere on the network. In most cases, you can simply install one new cable between the central connection point and the desired location of the new network device without moving or adding to a trunk cable or making the network unavailable for use by other stations. However, the star topology and its derivatives are also susceptible to bottlenecking and single-point failure; the latter is often remedied by providing a redundant backup of the hub node.
Also called a "hierarchical" or "star of stars" topology, tree topology is a combination of bus and star topologies. Nodes are connected in groups of star-configured workstations that branch out from a single "root," as shown in Figure 15. The root node usually controls the network and sometimes network traffic flow. This topology is easy to extend: when new users need to be added, it is simply a matter of adding a new hub. It also is easy to control because the root provides centralized management and monitoring. The principal disadvantage is obvious: when the entire network depends on one node, failure of that node will bring the whole network down. Also, the tree topology is difficult to configure, wire, and maintain, especially in extensive networks.
Figure 15: The tree topology is centrally controlled, making it easy to manage but highly vulnerable to single-point failure.
A topology gaining popularity in recent years is mesh topology. In a full mesh topology, each node is physically connected to every other node. Partial mesh topology uses fewer connections, and though less expensive is also less fault-tolerant. In a hybrid mesh the mesh is complete in some places but partial in others. Full mesh is generally utilized as a backbone where there are few nodes but a great need for fault tolerance, such as the backbone of a telecommunications company or ISP. Partial and hybrid meshes are usually found in peripheral networks connected to a full-mesh backbone.
The primary advantage of this topology is that it is highly fault tolerant: when one node fails, traffic can easily be diverted to other nodes. It is also not especially vulnerable to bottlenecks. On the other hand, as Figure 16 shows, full mesh topology can require inordinate amounts of cabling if there are more than just a few nodes. A full mesh is also complex and difficult to set up. In a partial or hybrid mesh there is a lack of symmetry—some nodes have more connections than others—which can cause problems with load balancing and traffic.
Figure 16: The full mesh, on top, is both highly complex and highly fault tolerant. The partial mesh sacrifices some fault tolerance in favor of increased simplicity.
Because the medium through which the signals are propagated (radio frequencies) has different properties than wires, wireless topologies differ greatly from wired topologies. The principles used in creating wireless networking solutions are based on the technology currently in use with cellular telephone systems.
Cellular technologies are often described in terms of their "generation": first, second, or third. The first generation is the analog cellular system, second-generation wireless is digital, and the third generation, which has yet to be developed, is often called UMTS: Universal Mobile Telecommunications System. This system is designed to provide digital, packet-switched, high-bandwidth, always-on service for everything from voice to video to data transfer. Once UMTS is implemented, it is hoped that it will be the only standard to which all cellular and wireless devices are built, thereby creating a universal wireless standard.
Literally at the center of any cellular technology is the cellular transceiver, an omnidirectional antenna whose range projects a circular "footprint." This footprint is the "cell" that gives cellular technology its name.
Cellular providers are allotted a set of frequencies within a specified area called a metropolitan statistical area (MSA) or a rural statistical area (RSA) (usually, two providers obtain rights to the same MSA or RSA). They divide their statistical areas into cells and place an antenna at the center of each. Ideally, the antennas are located such that their entire allotted statistical area is covered by cells. In this way, a cellular user can move throughout the provider's statistical area and always be in range of an antenna.
As the cellular user moves from one cell to another, the user's signal is transferred from one antenna to another in a process called "handing off." The handing-off process is governed by a mobile telephone switching office (MTSO)—the hub through which all cellular calls are routed. Figure 17 shows how handing off is accomplished. Cellular antennas continuously send out a control signal to whichever mobile devices are in their cells. When a cellular device initiates a transmission in Cell A, for example, the device receives the control signal and sends back another signal in reply. This signal is received by several nearby antennas, but to the antenna closest to it (in Cell A) the signal will be strongest. The MTSO will therefore route the cellular transmission through the antenna in Cell A; other antennas will ignore the signal. As the user moves from Cell A to Cell B, the MTSO will detect that the signal in Cell A is becoming weaker while in Cell B it is becoming stronger. When the user moves into Cell B, the MTSO assigns the device a new frequency and routes the transmission solely through the antenna in Cell B.
Figure 17: Hand-off from one cell to the next
In most wireless technologies, only one party can transmit a signal at a given frequency within a defined geographical location. With regard to cellular service, however, a person on a mobile call monopolizes their allocated frequency only within their current cell; someone in a cell across town can use that frequency at the same time. The concept of multiple users operating at the same frequency in the same geographic area is known as "frequency reuse."
For frequency reuse to be effective, every cell phone in the area must put out only enough power to reach the antenna of the cell that they are in. Too much power, and the signal would be picked up by unintended antennas in other cells; this could interfere with conversations in those cells being transmitted at the same frequency.
Cells using the same frequency, however, cannot be adjacent to each other. This is because a cellular caller on the border between two cells would put out just enough power to reach both the intended antenna in one cell and the antenna in the other. This would interfere with any conversations taking place at the same frequency in the unintended cell. Of course, in order for the cellular provider to allow more callers to use the same frequency to carry on separate conversations, the provider will want to place cells using that frequency as close together as possible. The number of cells of separation usually ranges from four to 21.
The term "air interface" describes the way the signal is modulated between the wireless device and the base station. Air interfaces generally use modulation schemes designed to increase the information throughput of the wireless system. Three principal air interfaces are Frequency Division Multiple Access (FDMA), Time Division Multiple Access (TDMA), and Code Division Multiple Access (CDMA).
FDMA is the oldest of these schemes and the least efficient. With FDMA, only one transmission is propagated over each channel at a time. The channel is dedicated to that one transmission regardless of whether data is being transmitted, and the channel is not available for another user until the device using it terminates the transmission. Figure 18 shows how FDMA allocates one channel for each user. This scheme is used by Advanced Mobile Phone System (AMPS), the analog cellular system in the United States, and Total Access Communication System (TACS), the analog cellular system in Europe. FDMA is for analog signals only and therefore can handle only voice transmissions, not data or fax.
Figure 18: FDMA channel allocation
TDMA is often called digital FDMA. It increases the number of transmissions per channel by taking advantage of the properties of digital technology. With digital transmissions, data is sent in discrete packets rather than continuous analog waves. TDMA assigns these packets a time slot on a frequency and alternates between transmissions so that more transmissions can be sent per channel. To send three telephone conversations on one channel, for example, TDMA sends first a packet from the first conversation, then a packet from the second, then from the third, then from the first, etc., as shown in Figure 19. By separating the conversations by time, TDMA can "simultaneously" transmit all three conversations over the same channel. The IS-54 and IS-136 implementations of TDMA allow three users to occupy the same channel at once, and there are implementations that allow six. In the future, TDMA will be able to accommodate up to 40 signals per channel. TDMA is used by GSM, the European digital standard, and PDC, the Japanese digital standard.
Figure 19: TDMA divides each channel into multiple time slots.
However, even though TDMA packs more transmissions into each channel, the time slots may go empty when data is not being sent, such as during the pauses in a conversation. An advanced implementation of TDMA, called Extended TDMA (ETDMA), assigns data to the time slots dynamically so that slots are always filled.
CDMA is DSSS technology. It attaches "pseudo-random code sequences" to each packet sent. The code is known only to the sender and the receiver. The signal is also spread across a range of frequencies, which allows many users to transmit across the same range at once, as shown in Figure 20. This spread-spectrum technology allows up to 20 times more transmissions per cell than with FDMA. CDMA is the predominant technology used by PCS networks in the United States.
Figure 20: CDMA spreads each transmission over a range of frequencies.
Wireless LANs (WLANs) are variations on wired LANs rather than new topologies, but the hardware used to create them is different. The most obvious hardware difference is the absence of wiring, which provides several advantages, not the least of which are lower cost and greater flexibility and mobility.
A wireless peer-to-peer network consists of two or more wireless-enabled devices such as PCs with WLAN cards or handheld devices that are in close proximity to one another. Figure 21 shows a wireless peer-to-peer network configuration.
Figure 21: The wireless peer-to-peer network consists of several wireless devices
A wireless topology that more closely mimics a wired LAN involves one or more APs or EPs. Figure 22 shows how an AP, wired to the LAN backbone, functions much like a hub or a switch.
Figure 22: Client and access point
The logical topology is the schema of the actual path the data follows within the physical topology. It differs from the physical topology in that not only does it show the location of network components, it also shows the path the data follows through these components as well as the direction of travel. It is used to enable network devices to transmit and receive data across the transmission media without interfering with each other.
Because the logical topology is associated with the path and direction of data, it is closely linked with the MAC methods in the media access layer of the OSI model. Specific MAC methods are required for each of the logical topologies in order to monitor and control data flow. These methods will be discussed in conjunction with the appropriate logical topology.
There are three basic logical topologies: logical bus, logical ring, and logical star (switching). Each of these topologies offers distinct advantages in specific situations. As you study the figures representing these topologies, remember that they illustrate a logical (electronic) rather than a physical connection scheme.
In the logical bus topology, transmissions (called frames) are broadcast simultaneously in every direction to every point on the transmission media. Every network station checks each frame to determine for whom it is intended. When the signal reaches any end point on the transmission media, it is absorbed (removed from the media) by an appropriate device. Removing the signal prevents it from being reflected back along the transmission media and interfering with subsequent transmissions.
On a logical bus network the transmission media are shared. To prevent transmission interference, only one station may transmit at a time. Thus, there must be a method for determining when each station is allowed to use the media. The methods used to control how data is sent on the network are the MAC methods that we discussed briefly in the "Data Transmission" section.
The MAC method most commonly used for a logical bus network is CSMA/CD. This MAC method is similar to the access scheme used on a telephone party line. When any station wants to send a transmission, it "listens" (carrier sense) to determine if another station is currently transmitting on the media. If another station is transmitting, the station that wants to transmit waits. When the media become free, the waiting station transmits. If two or more stations determine that the media are free and transmit simultaneously, there is a data "collision." All transmitting stations detect the collision and transmit a brief signal to inform all other stations that a collision has occurred. All stations then wait a random amount of time before attempting to retransmit.
A logical bus network may also use token passing for media access control. In this MAC method each network station is assigned a logical position in an ordered sequence with the last number of the sequence pointing back to the first (the logical order that the stations are assigned need not correspond to any physical order). A control frame called a "token" is used to control which station can use the media: a station can transmit only when in possession of the token. Furthermore, a station can have the token for a limited time only; it then must pass the token to the next station. The token starts at the first station in the predefined logical order. While the first station has the token, it transmits, polls stations, and receives responses until the allotted time expires; or, it passes the token when it no longer needs control of the media. The first station passes the token to the second station in the logical sequence. This sequential token passing continues nonstop while the network is running so that every station gets equal access to the transmission media.
The logical bus transmission scheme is used in combination with both the physical bus and physical star topology. The MAC method can vary from case to case. For example, while thin Ethernet and 10Base-T Ethernet use the logical bus transmission scheme, cable on thin Ethernet networks is laid out as a physical bus, and on 10Base-T networks as a physical star. Thin Ethernet (physical bus) and 10Base-T Ethernet (physical star), however, both use the CSMA/CD MAC method.
Figure 23 shows a thin Ethernet network (physical bus, logical bus), and Figure 24 shows a 10Base-T Ethernet network (physical star, logical bus). In both figures, notice that the network signal (shown by the arrows) emanates from the sending station and travels in all directions to all parts of the transmission media.
Figure 23: Thin Ethernet network (physical bus, logical bus)
Figure 24: 10Base-T Ethernet network (physical star, logical bus)
In the logical ring topology, frames are transmitted in one direction around a physical ring until they have passed every point on the transmission media. The logical ring must be used in combination with a physical ring topology such as the star-wired ring explained earlier. Each station on the physical ring receives the signal from the previous station and repeats the signal for the next station. When a station transmits data, it gives the data the address of another station on the ring. The data is circulated around the ring through each station's repeater until it reaches the station to which it is addressed and is copied. The receiving station adds an acknowledgment of receipt to the frame. The frame continues on around the ring until it returns to the station from which it was originally transmitted. This station reads the acknowledgment and removes the signal from the ring. Figure 25 shows how data would flow on a logical ring network with a star-wired ring physical topology.
Figure 25: Logical ring topology
Media access control for the logical ring topology is almost always based on a form of token passing. However, stations are not necessarily granted media access in the same order in which they receive frames on the physical ring. IBM's Token-Ring network is a logical ring network based on the star-wired ring physical topology.
Logical Star (Switching)
In the logical star topology, network switches are used to restrict transmissions to a specific part of the transmission medium. Transmission path restriction is the identifying characteristic of a logical star.
In its pure form, switching provides a dedicated line for each end station. When one station transmits a signal to another station on the same switch, the switch transmits the signal only on the two paths connecting the sending and receiving station. Figure 26 shows how data would be transmitted from one station to another if two stations were directly connected to the same switch.
Figure 26: Switching
Most switching technology adds switching capability to existing connection standards, incorporating the logical connection schemes of the existing standards, such as the MAC methods. For example, a 10Base-T Ethernet switch supports the Ethernet CSMA/CD MAC method.
Switches have built-in connection logic and a significant quantity of fast memory. They can simultaneously service all connected stations at full access speed. Thus, when you connect a station directly to a switch, you can increase the total throughput of your network—a significant performance advantage.
Switching illustrates that a logical topology consists of the total of the various aspects of the electronic connection scheme, not just the MAC method. By combining new switching capabilities with existing logical connection schemes, engineers create a new logical topology.
Multiple switches can be connected using one or more physical topologies. Switches can be used not only to connect individual stations, but to connect groups of network stations, known as segments. Thus, in many circumstances switching can be used to improve the performance of your network.
Connecting a Simple Network
Now that we have discussed the hardware pieces that make up a network and considered the difference between physical and logical topology, we will illustrate how to connect some hardware devices to form a simple network. Figure 27 shows some of the hardware items we have discussed, connected to form a basic computer network.
Figure 27: Various networking hardware connected to form a simple network
The network in this figure includes the following components: three computers connected with a 10Base-T hub by means of unshielded twisted-pair wiring, an Ethernet 10Base-T network adapter installed inside each of the nodes, and a laser printer that is connected to one of the nodes.
The node at the bottom center of the illustration is a network server, and it controls the network. The other two nodes are workstations. The workstations use the network under the control of the network server. One workstation is an IBM PC and the other is an Apple Macintosh computer.
The 10Base-T hub serves as a common connection point for the three computers. It also repeats network signals.
The lines between the different components of the network represent the transmission medium: twisted-pair wiring. This 10Base-T network is connected in a physical star, but it is based on a logical bus that uses a contention scheme for the workstations to gain access to the transmission medium.
The printer in this network is connected directly to the server by means of a parallel interface cable, which is a standard connection method. The server accepts print jobs from either workstation and sends the jobs through the cable to the printer. While this is the simplest way to enable both workstations to use the printer, there are other ways to connect printers to a network. You can, for example, attach them to a computer set up as a dedicated print server, or to a computer running special software by which it functions as both workstation and print server. Many high-end printers are now manufactured with an internal network adapter so that they can be attached directly to the transmission medium at any physical point in the network.Return to Primer Index | Next Section