24
Chapter 2 Network Topologies, Architectures, and the OSI Model In this chapter we present several fundamental networking concepts. We begin with an overview of network topologies, giving examples of the two most general designs on which most topologies are based: point-to-point and broadcast. Next, we examine the con- cept of network architecture. As part of this discussion we introduce the idea of "layering" and present the Open Systems Interconnect (OSI) networking model as an example of lay- ered architecture. We conclude the chapter with an introduction to TCP/IP and show how it relates to the OSI model. An outline of the material we discuss follows: Concept of Network Topologies (Questions 1-2) Point-to-Point Networks: Star, Loop, Tree (Questions 3-9) Broadcast Networks: Bus, Ring, Satellite (Questions 10-22) Multidrop Networks (Question 23) Circuit- and Packet-Switched Networks (Questions 24-32) Network Architectures (Questions 33-34) The OSI Model (Questions 35--41) Connection-Oriented and Connectionless Services (Questions 42-46) The TCP/IP Model (Questions 47-52) 1. What is a network topology? A network topology is the basic design of a computer network. It is very much like a road map. It details how key network components such as nodes and links are intercon- nected. A network's topology is also comparable to the blueprints of a new home in which components such as the electrical system, heating and air conditioning system, and plumb- ing are integrated into the overall design. 2. In which ways can nodes and links be interconnected? There are three general interconnection schemes: point-to-point, broadcast, and multi- drop. 31

Networking Explained || Network Topologies, Architectures, and the OSI Model

Embed Size (px)

Citation preview

Page 1: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2

Network Topologies, Architectures, and the OSI Model

In this chapter we present several fundamental networking concepts. We begin with an overview of network topologies, giving examples of the two most general designs on which most topologies are based: point-to-point and broadcast. Next, we examine the con- cept of network architecture. As part of this discussion we introduce the idea of "layering" and present the Open Systems Interconnect (OSI) networking model as an example of lay- ered architecture. We conclude the chapter with an introduction to TCP/IP and show how it relates to the OSI model. An outline of the material we discuss follows:

�9 Concept of Network Topologies (Questions 1-2)

�9 Point-to-Point Networks: Star, Loop, Tree (Questions 3-9)

�9 Broadcast Networks: Bus, Ring, Satellite (Questions 10-22)

�9 Multidrop Networks (Question 23)

�9 Circuit- and Packet-Switched Networks (Questions 24-32)

�9 Network Architectures (Questions 33-34)

�9 The OSI Model (Questions 35--41)

�9 Connection-Oriented and Connectionless Services (Questions 42-46)

�9 The TCP/IP Model (Questions 47-52)

1. What is a network topology?

A network topology is the basic design of a computer network. It is very much like a road map. It details how key network components such as nodes and links are intercon- nected. A network's topology is also comparable to the blueprints of a new home in which components such as the electrical system, heating and air conditioning system, and plumb- ing are integrated into the overall design.

2. In which ways can nodes and links be interconnected?

There are three general interconnection schemes: point-to-point, broadcast, and multi- drop.

31

Page 2: Networking Explained || Network Topologies, Architectures, and the OSI Model

32 Networking Explained, Second Edition

3. Start with point-to-point. What's that?

A point-to-point network consists of nodes that can only communicate with adjacent nodes. It's like looking into a telescope and seeing only one planet out the eyepiece.

4. What do you mean by adjacent nodes?

Adjacent nodes are nodes that are next to each other (Figure 2.1). Adjacency is typically expressed by stating the number of hops required for data to travel from the source node to the destination node. A hop is a connection to or from an intermediate node on the path from the source to the destination. Adjacent nodes are always one hop from each other. Thus, one hop implies two directly connected (line-of-sight) nodes. In a more complex form, a point-to-point network might consist of thousands of nodes connected to adjacent nodes, with these adjacent nodes connected to other adjacent nodes, and so on.

5. What happens if a node needs to communicate with a nonadjacent node?

It does so indirectly via other adjacent nodes. The source node first transmits a message to its adjacent node. This message is passed serially through each intervening node until it finally reaches the destination node. Passing data through an adjacent node to another node is typically called bridging or routing, depending on the passing technique used to transfer the information.

6. Which network topologies are based on the point-to-point design?

There are several. Three very common ones are star, loop, and tree.

7. What does a star topology look like?

A simple star configuration is shown in Figure 2.2(a). A key characteristic of a star is the presence of a central processing hub, which serves as a wire center for connecting nodes. In order for nodes to communicate with each other, all data must pass through the hub. Consequently, a hub represents a single source of failure. A typical star configuration is shown in Figure 2.2(b). This is a 10BASE-T network (a type of Ethemet) consisting of nodes connected to an "Ethemet switch" via unshielded twisted-pair cable (UTP). (10BASE-T networks, Ethemet switches, and UTP cable are discussed in detail in subse- quent chapters.)

FIGURE 2.1 Example of a point-to-point network design. A characteristic of point-to-point net- works is adjacencynnodes can only communicate with nodes they are next to. Thus, node A can only communicate with node B, and node C can only communicate with node B. If nodes A and C need to communicate with each other, they do so using node B, which is adjacent to both A and C.

Page 3: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 33

FIGURE 2.2 A simple star configuration (a) involves a wiring center (or hub) to which all nodes are connected and through which all data must pass. The hub represents a single source of failure because, if it fails, then all connected nodes will not be able to communicate. A typical hub configu- ration is shown in (b).

FIGURE 2.3 A loop design (a) is a modified star configuration. Instead of using a wiring hub, nodes are connected directly by dedicated wiring. If every node has a link to every other node, then we have a complete loop (b). (A complete loop is also called fully meshed.) Note that in a complete loop the number of links each node has is one fewer than the number of nodes in the network. Loops are more reliable than stars because the potential for a single source of network failure is removed.

8. W h a t does a loop look like?

A loop is a modified version of a star. In a loop, nodes are connected via dedicated wir-

ing instead of through a centralized hub. An example of a simple loop is shown in Figure

2.3(a). This involves only one connection between any two nodes. Note that a single link failure does not cause the entire network to fail. Thus, loops are more reliable than stars. A highly reliable and more expensive loop design involves each node being connected to

every other node. This is called a complete loop and is shown in Figure 2.3(b). Note that in

a complete loop every node is adjacent to every other node. A complete loop is also

referred to as a fully-meshed design.

Page 4: Networking Explained || Network Topologies, Architectures, and the OSI Model

34 Networking Explained, Second Edition

9. What about a tree?

A tree topology is a hierarchical configuration. It consists of a root node or hub that is

connected to second level nodes or hubs. These "level 2" devices are connected to "level

3" devices, which in turn are connected to level 4 devices, and so forth. A simple tree

topology is shown in Figure 2.4. One application of a tree topology is IEEE 802.12,

known as 100VG-AnyLAN, in which hubs are cascaded to form a hierarchical topology.

An example of this network is shown in Figure 2.5.

FIGURE 2.4 A simple tree topology consists of nodes interconnected in a hierarchical con- figuration.

FIGURE 2.5 An example of a two-level hierarchical topology in a 100VG-AnyLAN network. Each hub has at least one uplink port, which connects to a higher-level hub; every port can be used as a downlink port to connect to an end node or a lower-level hub.

Page 5: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 35

10. What about broadcast. What is it and what distinguishes it from point-to-point?

A broadcast design consists of nodes that share a single communications channel. In a point-to-point design nodes do not share a common channel but instead are directly con- nected to each other.

11. Is this like a telephone party line?

Yes it is, and you are dating yourself. In contrast to point-to-point design, data sent by one machine are received by all other nodes connected to the shared channel. Hosts receiv- ing a transmission check the destination address of the message to determine if it is intended for them. If not, they discard the message. Thus, only the destination node responds. As an illustration, consider a classroom setting with a teacher and 23 students. If the teacher asks one student, George, a question, all 23 students hear the question but only George responds. This is analogous to a broadcast network.

Compare this to point-to-point communication. Let us now assume that one student, Patti, wants to tell her friend, Janie, who is sitting two rows over and three seats behind Patti, to wait for her after class. To get this message to Janie, Patti turns to John who is sit- ting behind her and says, "Tell Janie to wait for me after class." John turns to the person behind him and says, "Tell Janie to wait for Patti after class." This message continues being passed from one person to the next until it finally reaches Janie. This is an example of a point-to-point design.

12. Since all nodes hear a transmission in a broadcast design, is it possible for a node to send the same data to more than one node during the same transmission?

Yes. In fact, there are three different types of messages. The first is a unicast message, which is destined to only one recipient. The second is a multicast message, which is des- tined to a group of recipients. It is important to note that a node "knows" that it is in a mul- ticast group by its networking software "telling" it to listen to the multicast messages for the group. In many cases, the sending system to the multicast group does not know which nodes are actually members of the group. The third is a broadcast message, which is des- fined to all hosts connected to the network. A broadcast message is a special multicast message. (Note: In IPv6, discussed in Chapter 3, there is also the concept of anycast.)

13. How is one type of broadcast message distinguished from another?

This is protocol-dependent. Some protocol suites do not use broadcast and only use multicast. Others do not use multicast and use broadcast for group addressing needs. Would you like an example?

14. Yes. Tell me how it is done in Ethernet/802.3 networks.

An Ethcmct/[EEE 802.3 address consists of 48 bits (eight bytes) that are represented as 12 hexadecimal digits (0-9, A-F) and partitioned into six groups of two. For example, 08:00:20:01:D6:2A is a valid Ethemet/802.3 address. If the second hexadecimal digit (from the left) of a destination address is 0 or an even digit (2, 4, 6, 8, A, C, E), then the message is unicast. Thus, 08:00:20:01:D6:2A is a unicast address because its second digit

Page 6: Networking Explained || Network Topologies, Architectures, and the OSI Model

36 Networking Explained, Second Edition

is 8, which is even. If this second hexadecimal digit is odd (1, 3, 5, 7, 9, B, D, F), then the message is multicast. Ethernet broadcast messages, which are special multicast messages, use the address FF:FF:FF:FF:FF:FE See Box 5.1 and Appendix A for additional informa- tion about Ethernet/802.3 addresses.

15. Since it is possible for a node to send the same data to more than one node during the same transmission, is it also possible for more than one node to send data at the same time?

Well, they can try but they are not going to be successful. Since all nodes share the same communication channel, they must contend for the channel when they transmit. Thus, broad- cast networks promote the concept of contention and hence require some sort of method for governing those cases when two or more nodes attempt to transmit data at the same time.

16. What sorts of protocols are there for resolving such squabbles among nodes?

There are many. We provide an overview of some of the more popular ones in Chapter 5 and describe them in more detail in subsequent chapters.

17. OK. Now that I know a little bit about broadcast designs, which topologies are based on this design?

Broadcast networks employ several topologies. They are bus, ring, and satellite.

18. Could you give an illustration of each of these broadcast-based designs?

Yes. A typical bus configuration is shown in Figure 2.6, a ring is shown in Figure 2.7, and a satellite is shown in Figure 2.8.

19. It 's quite apparent from Figure 2.6 how a bus is a broadcast topology. It clearly shows all the nodes connected to the same channel. What about the ring and sat- ellite configurations? How are they broadcast-based?

If you look closely at Figure 2.7 you will notice that all nodes are connected to the same ring, which serves as the shared medium. Remember, a broadcast design means all nodes share a single communications channel, and that messages sent by one node are received by all others connected to this channel.

20. How is a ring different from a loop, which is shown in Figure 2.3, and which is point-to-point? They almost look the same.

You are right. They do look similar. Looks are deceiving, though. The distinction can be made by looking at things from a logical and physical perspective. In a ring configura- tion, all nodes are connected to the same ring, which serves as the shared medium. Ring- based networks can be designed physically as a star (Figure 2.7a), or as a simple loop (Figure 2.7b). The star design is formally referred to as a logical ring over a physical star, and the simple loop design is formally referred to as a logical ring over a physical ring.

Page 7: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 37

FIGURE 2.6 Typical bus configuration. An example of a topology based on a broadcast design.

FIGURE 2.7 Ring-based networks represent a topology based on a broadcast design. Rings can be configured as (a) a logical ring over a physical star or (b) as a logical ring over a physical ring.

FIGURE 2.8 Typical satellite configuration. Nodes use an antenna to send and receive data.

The key difference between a ring-based "loop" (Figure 2.7b) and a point-to-point- based loop (Figure 2.3a) is that the ring-based loop is a broadcast system and hence, nodes share a single communications channel. In a point-to-point system, however, we have the concept of adjacency and hence nodes can only communicate with the node next to them. If you examine the loop topology in Figure 2.3a, you will notice that the nodes do not share a single channel. There is a specific, dedicated link between any two nodes.

Page 8: Networking Explained || Network Topologies, Architectures, and the OSI Model

38 Networking Explained, Second Edition

21. How does a ring work?

In a classic ring topology, messages are passed from node to node around the ring. The direction of rotation can either be clockwise or counterclockwise (or both), depending on the technology. As is the case for the bus topology and for all broadcast systems, some method must govern simultaneous ring access. Note that although data are passed from node to node, this is not a point-to-point topology because all the nodes share the same com- munications channel. Thus, logically, the topology of a ring involves all nodes sharing the same communications channel. Physically, though, the communications are point-to-point.

22. OK. Now what about satellite networks? What makes them broadcast systems?

In a satellite communication system, data transmissions from a land-based antenna to the satellite are generally point-to-point. However, all nodes that are part of the network can receive the satellite's downlink transmissions. (A downlink is the communication link from the orbiting satellite to one or more ground stations.) This classifies them as broadcast sys- tems. For example, many schools in the United States have satellite downlink capabilities. Whenever an educational program is broadcast via a satellite system, school sites wishing to receive this transmission simply tune their receivers to the proper frequency.

23. What's a multidrop network design?

In some types of networks, especially factory networks and those used to control real- time activities like power company networks, a particular design concept called a multidrop network is frequently used. Multidrop networks typically employ a master-slave concept with one node being assigned the network master and all other nodes being slaves. In this setting, the master controls the network functions and the slaves request network access from the master. Nodes are connected to a common cable plant similar to a bus design, but unlike bus networks multidrop nodes are assigned a specific number for communication purposes. This number also is used to establish priority of when a system is permitted to communicate with the master control system. This allows total control over the prioritiza- tion of traffic on the network as well as total control over the use of the network. Multidrop networks are popular only in factories because they are not terribly fast networks and would not work well in offices where users might want to share large disk drives and applications. They are typically used for command-and-control operations and some light data transfer of material information or tracking information (like bar codes). Multidrop networks are also often seen on older (circa early 1980s), legacy dumb terminal networks to reduce network costs. An illustration of a multidrop design is given in Figure 2.9.

24. What other network classifications are there?

In addition to geographical area and topology, networks also can be classified by the type of communications path they use and the manner in which data are transmitted across this path. Two particular classifications are circuit-switched and packet-switched. Switched networks involve a partially- or fully-meshed topology (see Figure 2.3) and use special network devices called switches to interconnect the links between source and des- tination nodes.

Page 9: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 39

FIGURE 2.9 In a multidrop network topology, all nodes are connected to a common medium. Us- ing a master-slave concept, where one node is the designated master and all other nodes are slaves, the master controls all network functions and the slaves request network access from the master. In this illustration, node A is the master, and nodes B, C, D, and E are slaves.

25. What's the difference between the two types of switched network?

In a circuit-switched network, a dedicated, physical circuit is first established between the source and destination nodes before any data transmission takes place. Furthermore, this circuit remains in place for the duration of a transmission. The public telephone sys- tem, known formally as the Public Switched Telephone Network (PSTN), is an example of a circuit-switched network. When we place a telephone call, a direct physical communica- tions path is established between our telephone set and the receiver's telephone set. This path is a point-to-point connection that interconnects the telephone company's switches, which are located throughout the telephone network. Once established, the circuit is dedi- cated exclusively to the current transmission. After the transmission is completed, this dedicated circuit is then released and made available for another communication transmis- sion. Thus, circuit-switching promotes link sharing since the same circuits can be used for different transmissions, although not at the same time.

In a packet-switched network, messages are first partitioned into smaller units called packets, which are then sent to the destination node one at a time via intermediate switches. A packet represents the smallest unit of data that can be transferred within a given network. Each packet carries the destination node's address as well as a sequence number. When a packet arrives at an intermediate switch, the switch examines the packet's destination address to determine which path the packet should take to the next switch. This switching technique in which data are stored on one node of a point-to-point link and then forwarded to the next node repeatedly en route to the destination node is called store-and- forward. The concept of a store-and-forward transmission requires that the entire contents of a transmitted message (or packet) be received by each intermediate node before it is for- warded to a succeeding node.

26. So circuit-switching sets up a link between the two end nodes and then transmits the data across this link. Packet-switching, on the other hand, partitions the data into packets and then transmits the packets. Is this right so far?

Yes.

Page 10: Networking Explained || Network Topologies, Architectures, and the OSI Model

40 Networking Explained, Second Edition

27. If this is the case, then on what circuit are the packets transmitted in a packet- switching environment? How can the packets be transmitted if there is no link established between the two end nodes as there is with circuit-switching?

You' re right, and something is missing. We need to mention that packet-switched net- works also promote the concept of link sharing, which can be effected in two ways: Link sharing can be accomplished by using virtual circuits or a datagram transport scheme. In virtual-circuit packet-switching, instead of using a dedicated, physical circuit for every node-to-node communication (as is the case with circuit-switching), nodes share a com- munications channel via a virtual circuit. A virtual circuit is a logical communications path instead of a physical one. That is, it is a nondedicated logical connection through a shared medium that gives the high-level user the appearance of a dedicated, direct physical connection from the source node to the destination node. In this transport mechanism, individual packets follow the same communications path in sequence as if they were trav- eling along a dedicated circuit. The difference is that links within this circuit also can be used for other transmissions at the same time.

28. Wait a minute. Before we go any further, I need an example.

OK. Let's compare circuit-switching to virtual-circuit packet-switching. As an illustra- tion of circuit switching, consider the fully meshed, point-to-point design shown in Figure 2.3(b). If a communications path between nodes 1 and 3 is established by interconnecting links 1-2-4-3, then these links collectively form a dedicated, physical circuit between the two nodes, and this circuit is monopolized by these two nodes for the duration of their communications. However, once their transmission ends, the circuit is released and the individual links that made up this circuit can now be used to construct a new dedicated cir- cuit between two other nodes (e.g., 4-1-2). In a virtual-circuit packet-switching environ- ment, node 1 can be communicating with node 3 along the 1-2-4-3 path, and at the same time, node 4 can be communicating with node 2 along the 4-1-2 path. Both paths represent virtual circuits that simultaneously use the 1-2 link. Contrast this with circuit-switching in which the circuit is dedicated and the individual links comprising the circuit cannot be used simultaneously to construct any other circuit.

29. OhX I get it. So that's the difference between a dedicated link versus link sharing. OK. Now what about this datagram transport scheme you mentioned?

As we stated earlier, in addition to virtual circuits, packet-switched networks can also use a datagram transport mechanism for path selection. In datagram packet-switching, packets are transmitted independently of one another at any time. Thus, it is possible for different packets from the same message to be transported across different communica- tions paths. Furthermore, the packets are not necessarily transmitted in a specific order, which implies that the specified destination node is responsible for reassembling them in the correct order. (This is why packets also carry sequence numbers.) Most modem-day computer networks, including the Intemet, are packet-switched.

Page 11: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 41

30. How are virtual circuits formed?

A virtual circuit is created by multiplexing a physical link so that it can be shared by multiple network programs or data transmissions. We will not go into the details of multi- plexing here. If you're really interested in it at this stage, though, feel free to jump ahead to Chapter 4 where it is discussed.

Multiplexing is extremely valuable for providing low-cost communications capabili- ties because it is very expensive to provide dedicated links for every data transmission, as in circuit-switched networks. A definition of virtual has been memorably coined in this way: "If you can see it and touch it, it's physical; if you can see it but can't touch it, it's virtual; if you can't see it and can't touch it, it's gone."

31. So, is the main difference between circuit-switching and packet-switching the type of link that is used, namely, one is dedicated and the other is virtual?

Not quite, virtual reader. The main difference between circuit- and packet-switched networks is the use of bandwidth, which is the maximum capacity of a communications channel. For example, a circuit might have a data transmission capacity of 100 megabits per second. We will discuss the concept of bandwidth in more detail in Chapter 4. For the present, though, just understand that in a circuit-switched network a circuit's performance is predetermined and fixed. This means that bandwidth is allocated in advance and guaran- teed for the entire transmission. Once a circuit is established, the full capacity of the cir- cuit is available and the capacity of the circuit will never be reduced due to other network activity. This advantage of circuit-switched networks also gives rise to a disadvantage. Specifically, circuit costs are independent of the amount of data being transmitted; there- fore, any unused bandwidth is wasted.

On the other hand, packet-switched networks acquire and release bandwidth dynami- cally as needed. One major advantage is that several communications can occur between nodes concurrently using the same channel. Again, this advantage becomes a disadvantage when, as packet-switched networks become overloaded with more traffic, delays and con- gestion are introduced. Nevertheless, packet-switched networks are cheaper and offer bet- ter performance than circuit-switched networks. Furthermore, given recent developments in high-speed switching hardware, the channel capacity issue has eased a bit. Table 2.1 provides a summary of the differences between circuit- and packet-switching.

32. Given the advantages and disadvantages of circuit- and packet-switched net- works, why not combine the two so you could have the best of both worlds?

You are referring to hybrid switching, which combines the principles of both circuit- and packet-switching. This technique first partitions a message into packets (packet- switching) and transmits each packet via a dedicated circuit (circuit-switching). As soon as a packet is ready for transmission, a circuit meeting appropriate bandwidth requirements is established between the sending and receiving nodes. When the packet reaches its destina- tion, the circuit is terminated (i.e., it is "tom down" using telephone terminology) so that it can be used again. This scenario has many advantages but it also requires extremely fast circuit-switching equipment.

Page 12: Networking Explained || Network Topologies, Architectures, and the OSI Model

42 Networking Explained, Second Edition

TABLE 2.1 Circuit-Switching versus Packet-Switching

Circuit-Switched Packet-Switched

1. Bandwidth is allocated in advance and is guaranteed for the entire transmission.

2. Once circuit is established, the full capacity of the circuit is available for use, and the capacity of the circuit will never be reduced due to other network activity.

3. Circuit costs are independent of the amount of data being transmitted and hence any unused bandwidth is wasted.

1. Bandwidth is acquired and released dynamically on an as-needed basis.

2. Several communications can occur between nodes concurrently using the virtual links over the same physical channel.

3. As packet-switched networks become overloaded with more traffic, delays and congestion are introduced.

4. Packet-switched networks are more cost- effective and offer better performance than circuit-switched networks.

Thus far in this chapter, we have examined various network topologies. Many of the designs we discussed will be reintroduced in later chapters that deal with specific networking tech- nologies such as Ethernet and token ring. The concepts of circuit-switching, packet-switch- ing, and multiplexing will also be re-examined in more detail in subsequent chapters. We now focus our attention to network architectures and the OSI and TCP/IP models.

33. What is a network architecture?

Network architecture is a formal, logical structure that defines how network devices and software interact and function. It defines communication protocols, message formats, and standards required for interoperability. New hardware or software products created within a specific architecture are generally compatible with other products created within the same architecture.

34. Who creates or designs network architectures?

Network architectures are designed by standards organizations and manufacturers. For example, IBM designed the Systems Network Architecture (SNA), the former Digital Equipment Corporation (DEC) designed the Digital Network Architecture (DNA), and the International Organization for Standardization, which is always abbreviated "ISO," designed the Open Systems Interconnect (OSI) architecture.

35. I've heard of these architectures before and am familiar with SNA and DNA. Tell me, what's the big deal about OSI?

Well, OSI really isn't much of a big deal anymore, although it did once have a very pronounced role in the networking community. The genesis of OSI can be thought of as follows: Once upon a t ime--a long, long time ago (circa 1970s)--there was no such thing as a network architecture. Companies designed rather rude, crude, and socially unaccept- able proprietary software and hardware communications products without any consider-

Page 13: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 43

ation to the implementation of a coherent architecture--the long-term technical effect of decisions made when something is constructed. Eventually, issues of interoperability and design began to emerge as new networks were being developed.

To address these issues, and to accommodate interconnection of the various propri- etary and heterogeneous networks, ISO developed in 1978 a seven-layer architecture and reference model intended to serve as the foundation for future standards activities. The resulting model was formally called the Basic Reference Model for Open Systems Inter- connection, or the OSI model for short. The OSI model provides a detailed set of standards for describing a network; it is a framework for the development of network protocol stan- dards. The OSI model formally defines and codifies the concept of layered network archi- tecture. It uses well-defined operationally descriptive layers that describe what happens at each stage in the processing of data for transmission.

36. What's so great about layers?

Networks are nontrivial systems. Given a network's complex nature, it is extremely dif- ficult to design an architecture that (a) has a high degree of connectivity, (b) is reliable, and (c) is easy to implement, use, and modify. Layers help reduce the design complexity of a network. By organizing a network's functions into a series of hierarchical layers, the design of a network is greatly simplified. For example, a layered approach enables the functions and services of one layer to be completely independent of and isolated from other layers. Thus, a layered approach effectively screens from other layers the actual implementation details of the services one layer is offering to another layer. This allows us to change a layer's capabilities without significantly modifying the entire architecture. So as new tech- nologies become available for one layer, they can be implemented without affecting the operation of the other layers. In theory, a layer can be completely removed, dramatically changed, and reinserted without affecting the operation of the layers above or below it.

37. Is this like modular programming?

You bet it is. Just as large computer programs are partitioned into separate, indepen- dent program modules, layers partition a network architecture into separate, independent components. Each layer is responsible for performing a specific set of functions and for providing a specific set of services. Specific protocols define both the services and the manner in which these services are provided. Another analogy is an automobile, which comprises several independent components, including the electrical, braking, cooling, and ignition systems. Working on one component does not affect another component. Thus, if we need to replace our car's brakes, we do not have to be concerned with how it will impact the car's ignition system.

38. What are the layers of the OSI model?

The layers of the OSI model are (from top to bottom): application, presentation, session, transport, network, data link, and physical. These layers are numbered in descending order from seven to one and define the communication capabilities needed to effect communica- tion between any two devices. Figures 2.10 and 2.11 provide additional information about OSI layers.

Page 14: Networking Explained || Network Topologies, Architectures, and the OSI Model

44 Networking Explained, Second Edition

Application (7) �9 Consists of protocols that define specific user-oriented applications such as

e-mail, file transfers, and virtual terminal. �9 Examples include FTAM (File Transfer, Access, and Management) for remote file

handling, X.400 (for e-mail), and CMIP (Common Management Information Protocol) for network management.

Presentation (6) �9 Provides data formats, translations, and code conversions. �9 Concerned with syntax and semantics of data being transmitted. �9 Encodes messages in a format that is suitable for electronic transmission. �9 Performs data compression and encryption. �9 Receives/formats message from application layer and passes it to session layer. �9 In practice, this layer is usually incorporated within the application layer.

Session (5) �9 Provides coordination between communicating processes between nodes. �9 Responsible for enforcing the rules of dialog (e.g., Does a connection permit half-

duplex or full-duplex communication?), synchronizing the flow of data, and reestablishing a connection in the event a failure occurs.

�9 Examples include AppleTalk Data Stream Protocol for reliable data transfer between two nodes, NetBEUI (an extension of NetBIOS), and Printer Access Protocol for accessing a PostScript printer in an AppleTalk network.

�9 Uses the presentation layer above it and the transport layer below it.

Transport (4) �9 Provides error-free delivery of data. �9 Accepts data from the session layer, partitions data into smaller packets if

necessary, passes these packets to the network layer, and ensures that packets arrive completely and correctly at their destination.

�9 Examples involve varying classes of the OSI Transfer ProtocolmTPx, where x = {0, 1,2, 3, 4}. Each class describes a specific level of service quality such as whether a transmission provides for error detection or correction, or if the service is connection-oriented or connectionless.

Network (3) �9 Provides end-to-end routing or switching, which establishes a connection for the

transparent delivery of data. �9 Addresses and resolves all inherent problems related to data transmission

between heterogeneous networks. �9 Uses the transport layer above it and the data link layer below it. �9 Formatted messages are referred to as packets.

Data Link (2) �9 Responsible for end-to-end data transfer across a physical link. �9 Provides error detection, "framing," and flow control. �9 Resolves problems due to damaged, lost, or duplicate frames. �9 Formatted messages are referred to as frames.

Physical (1) �9 Responsible for transmitting raw bits over a link; it moves energy. �9 Accepts frames from the data link layer and translates the bit stream into signals

on the physical medium, which lies below it. �9 Concerned with issues such as the type of wire being used, the type of connector

(i.e., interface) used to connect a device to the medium, and signaling scheme.

FIGURE 2.10 Summary of the OSI layers and functions.

Page 15: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 45

Software

Hardware

Application (7) Provides user-application services and

procedures

Presentation (6) Structures data in a mutually agreed

format; concerned with issues such as how to code and format data; includes data

encryption

Session (5) Controls process communications;

responsible for segmenting, buffering, and synchronization

Transport (4) Provides end-to-end control; responsible for

partitioning and reassembling messages

Network (3) Provides routing services for transferring

data across the network; performs network management, packet formatting

Data Link (2) Organizes data into frames; provides flow initialization, flow control, link termination,

and error control

Physical (1) Transfers bits across link, defines physical

characteristics of media

Application or Service-Oriented

Layers

Delivery and Verification

Services

Communication or Network-

Oriented Layers

FIGURE 2.11 Another view of the OSI layers. Source: Adapted from Conrad, 1988b.

39. Why seven layers? Why not 5 or 20?

Seven is an interesting number. For example, there are the seven wonders of the world, the universe was created in seven days (or so we are told), there is supposed to be something mystical about being born a seventh son, seven is presumably a lucky number (although there is the slogan "seven years bad luck"), and major league baseball, basketball, and hockey have seven-game playoff series. So, why not seven? Seriously, though, the selection of seven layers was based on specific principles and design criteria. Two specific layers- related criteria were (a) a layer should be created only when a different level of abstraction is required, and (b) the number of layers should be large enough to keep distinct functions separate but not too large such that the architecture becomes too cumbersome or unman- ageable. Following is the logic used in creating each layer based on the established criteria:

�9 The designers first agreed that a separation between applications and media was needed. This established two layersman application layer and a media layer.

�9 If true global networking services were going to be provided, then the media layer required further refinement. Specifically, one layer was needed to specify the

Page 16: Networking Explained || Network Topologies, Architectures, and the OSI Model

46 Networking Explained, Second Edition

media and corresponding signaling. This resulted in the creation of the physical layer. Another layer was needed to specify the operation of a single link. This is the role of the data link layer. Finally, a network layer was needed to provide end- to-end network service that specified the operation of end nodes as well as inter- mediate nodes across different links and topologies. Thus, the original media layer was supplanted by the physical, data link, and network layers. This now gives us four layers: the three "media-related" layers and the application layer.

The media-related layers effectively addressed technology issues and hence were consid- ered technology-dependent and mutually exclusive of network applications. With these issues aside, the designers then focused on the actual data transmission relative to various aspects of an application. This resulted in three additional layers:

�9 First, the architecture designers agreed that the sending and receiving nodes of a data transmission should have some assurance that a transmitted message was received correctly and that the sending node was informed of this. Thus, what was needed was a mechanism for reliable, end-to-end transmissions. This led to the establishment of the transport layer, which effectively provides a quality of service. The inclusion of a transport layer brings the count to five layers.

�9 Second, the designers decided there should be a mechanism for the sending and receiving nodes to establish a dialogue about how to exchange messages. For example, the end nodes should be able to communicate with each other about how a message was going to be delimited (i.e., where it starts and ends) and how the virtual session between the two end processes was going to be estab- lished, monitored, and terminated. The end nodes also needed to exchange mes- sage synchronization information, which provides checkpoints throughout the transmission so that the two nodes are cognizant of buffer space for message storage as well as when the entire message has been received. This led to the session layer which brings the total number of layers to six.

�9 Finally, the designers recognized that application messages could be represented in many different ways depending on the circumstances in which a message was created and transmitted. For example, the encoding of a message might be in ASCII or EBCDIC, the message might be in integer format or floating point, an end-of-line character could be a carriage return (CR) or a carriage return fol- lowed by a line feed (CR/LF), or the data might be compressed, which reduces the number of bits needed to be transmitted. This led to the development of the presentation layer (which probably should have been more appropriately named representation or encoding layer). In short, the presentation layer is concerned with bit-pattern representation during transfer. The presentation layer was later modified to incorporate a standard language for specifying abstract syntaxes along with a set of encoding rules associated with use of that language. Thus, the presentation layer on the sending node translates the application data format it receives from the application layer into a common language format. The pre- sentation layer on the receiving node then translates this common format to the application format, which is then passed to the application layer.

Page 17: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 47

40. All of this certainly makes sense, but what about network management or secu- rity? Shouldn't there have been separate layers for them?

Good question. The issues of management and security did not escape the designers of the OSI model. For example, they felt that management did not warrant a separate layer because the functions of network management (e.g., traffic monitoring) were considered applications themselves. As for security, the designers decided that security features should be provided by the layers and hence incorporated various aspects of security into most of them.

41. How do layers communicate with each other?

Look at Figures 2.12 and 2.13. The data transmission of a message begins on the sending node. An application process first creates a message, attaches an application header to it, and then passes the message to the presentation layer. The presentation layer, if necessary, trans- forms the message into a different format (e.g., translating the data from the format sent by the application layer into a common format), attaches a header to it, and passes it to the session layer. This process is then repeated from layer to layer. When the message reaches the phys- ical layer, it is then transmitted to the destination node. In some cases a "trailer" is also appended to a message. For example, at layer 2 a trailer is added to facilitate frame synchro- nization. This is covered in more detail in Chapter 5. This process is then reversed on the receiving node with each layer's header (or trailer) being stripped off one by one as the mes- sage ascends the layers. Note from Figure 2.12 that at layer 3 a message is called a packet, at layer 2 it's called a frame, and at layer 1 it's referred to as bits.

FIGURE 2.12 The OSI layering process begins at the application layer of the source machine where a message is created by an application program. This message moves down through the layers until it reaches layer 1. Underlying layer 1 is the actual physical medium. Data are then transmitted across this medium to the receiving host, where the information works its way up through the layers. As messages move down the layers, they are encapsulated with headers that are germane to a specific layer. These headers are removed as the data are passed upward through corresponding layers at the receiving host.

Page 18: Networking Explained || Network Topologies, Architectures, and the OSI Model

48 Networking Explained, Second Edition

I Data Message[

I Data Message I - - ~ - - ~ - ' ~

p - - . . q 41

01

FIGURE 2.13 How layers work. Each layer "envelops" the data with its protocol. Each layer has a corresponding layer on the remote (destination) node, which is called a peer.

Additionally, each layer consists of two parts: a service definition, which defines the type of service a layer provides, and a protocol specification that details the rules governing the implementation of a particular service. Lower layers provide services to upper layers.

42. What kinds of services do layers provide to each other?

There are two different types of services: connection-oriented and connectionless. Some layers have an additional type of service called multiplexing, but this does not neces- sarily transcend all layers of the architecture. Services are available at service access points (SAPs), with each SAP having a corresponding address. (Note: In UNIX, a SAP is called a socket, and a SAP address is a socket number.)

43. Define connection-oriented service.

Connection-oriented implies that prior to the transfer of data a physical (and virtual) link is established between the sending and receiving nodes. This link remains in effect for the duration of the session. After the session is completed, the link is removed. Character- istics of a connection-oriented service include: wasted bandwidth, because the link must remain established even during idle periods of a transmission; a high potential for a hung network, since there is always a possibility that a link will not be terminated; and (on the bright side) guaranteed sequential arrival of packets at the destination node.

Page 19: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 49

The telephone system is an example of connection-oriented service. You establish a connection (you dial a number); you transfer data over this circuit when the connection is made (you begin talking when the receiver is picked up); communication occurs in the proper sequence (words and sentences are received in the correct order); and you release the connection at the conclusion of the transfer (you hang up the phone, which frees the circuit). Note also the issues of wasted bandwidth and a hung network. If a telephone con- nection has been made but no one is talking, bandwidth is wasted because the circuit is established but not being used. Anyone trying to contact your house during this period of silence would be greeted by a busy signalma "hung" connection.

44. What about connectionless service?

Connectionless service differs from connection-oriented service in that no physical link is established between sending and receiving nodes prior to data transmission. Instead, a message is partitioned into packets and routed through the network. Each packet is independent of the other packets that carry parts of the message and hence must carry a destination address. Note that addressing is not necessary for connection-oriented service because a physical, dedicated link is established between sending and receiving nodes before transmitting data. Packets can arrive out of order. Think of the post office as provid- ing connectionless service. If you send someone five separate letters numbered one through five, you must place the recipient's address on each letter. Once mailed, the letters do not necessarily follow exactly the same delivery route, and it is possible for the recipi- ent to receive the letters out of sequence (e.g., letter three is received before letter two).

Connectionless service is also either reliable or unreliable. Unreliable service requires no acknowledgment of receipt of data from the receiving node to the sending node. This is called a datagram service. Reliable service requires an acknowledgment. This is called an acknowledged datagram service. Using our post office metaphor, these services compare with mailing a "regular" letter versus mailing a registered letter with a return receipt request.

45. Can you sum all this up with an example?

We could, although the best and most practical example--the Internet--requires addi- tional information that has not yet been presented. For example, the Internet is based on the TCP/IP protocol suite, part of which is discussed in Chapter 3. Also, we have not yet presented detailed discussions of some of the OSI layers, such as the data link and network layers, which are presented in Chapters 5 and 7, respectively. Since you asked, we will give you an example but without a lot of elaboration.

46. Go for it. I promise not to ask for additional information. I just want to try to piece all of this together conceptually.

OK. Let's assume we want to send a message across the Internet. Before doing so, though, we should first provide you with some information about addressing methodology. Three different addresses are needed to send a message from one node to another. The first address is the hardware address, which uniquely identifies each node. Hardware addresses are provided by the data link layer. The second address is the network address, which

Page 20: Networking Explained || Network Topologies, Architectures, and the OSI Model

50 Networking Explained, Second Edition

identifies the network to which a node is connected. In TCP/IE this is called an Internet address or an IP (for Internet Protocol) address. Network addresses operate at the network layer. Each network node that is part of the Intemet has a unique IP address. (Note: IP addresses do more than simply identify the network. See Chapter 3 for additional informa- tion about IE) The third address is called the port address, which uniquely identifies a spe- cific user application such as e-mail. All network applications have corresponding identifiers called port numbers. To send a message from one node to another, a message is first created at the application layer. It undergoes whatever formatting is required as it descends the layers. When the message reaches the network layer, a network address is assigned to the message. This network address identifies the specific network to which the destination host is connected. Depending on the protocol, this service is either connection- less or connection-oriented. For example, TELNET and SMTP are connection-oriented services. The network layer determines the path the message must follow to reach the des- tination node. It also encapsulates packets into IP datagrams and passes them to the data link layer. At the data link layer, the destination node's hardware address is added to the packet. This address uniquely identifies the location of the destination node within the des- tination network. The data link layer, among other tasks, also formats the packet into frames, which are like packets but exist at a lower level and checks, the integrity of each frame (see Chapter 5). Frames are then passed to the physical layer, which places them on the medium for transmission.

47. OK. Enough of OSI. How does TCP/IP compare to OSI?

TCP/IP's development preceded the OSI model by several years. Both had similar design goals, however, to fill a need for interoperability among heterogeneous computer systems. Unlike OSI, TCP/IP was never intended to be an international standard. It was developed to satisfy the need to interconnect various United States Department of Defense projects, including computer networks, and to allow for the addition of dissimilar machines to the networks in a systematic, standardized manner.

48. Does TCP/IP also have seven layers like OSI?

No. As a pre-OSI protocol architecture, it was not designed specifically with layers the way the OSI model was designed and it does not fit neatly into the seven layers of the OSI model. However, we can envision TCP/IP's layers as similar to the OSI layers since many of TCP/IP's functions are similar to those of the OSI model.

49. So what are the layers?

There is no universal agreement on the description of TCP/IP as a layered model. It is frequently described as either a four- or five-layered model depending on an author's per- spective. For our purposes, we elect to describe TCP/IP as a four-layered architecture, which is shown in Figure 2.14. Note that a five-layered TCP/IP model maintains OSI's physical and data link layers as separate levels instead of combining them into a single layer as shown in Figure 2.14. In this scenario, TCP/IP's first layer is called the physical layer and its second layer is called the network access layer.

Page 21: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 51

OSI Layers TCP/IP Layers

Application

Presentation Application

Session

Transport

Network

Data Link

Physical

m

i Host-to-Host Transport

i Internet

Network Interface

TCP/IP's application layer corresponds to OSl's application, presentation, and session layers

TCP/IP's host-to-host transport layer corresponds to osrs transport layer TCP/IP's Internet layer corresponds to OSl's network layer

TCP/IP's network interface layer corresponds to OSl's data link and physical layers

FIGURE 2.14 A comparison of the OSI and TCP/IP layers.

50. Are TCP/IP's layers similar to the corresponding OSI layers in terms of the func- tions and services they provide?

Yes. The TCP/IP application layer serves as the communication interface for users by providing specific application services to the user such as remote terminal login (i.e., vir- tual terminal), file transfer, and e-mail. Corresponding application protocols include TEL- NET, FTP, and SMTP. The TCP/IP host-to-host transport layer (known simply as the transport layer) is responsible for end-to-end data delivery. This layer is defined by two protocols: the Transmission Control Protocol (TCP) and User Datagram Protocol (UDP). A brief description of each follows. A summary description of the TCP/IP model is also given in Figures 2.15 and 2.16.

�9 TCP. This is the TCP of TCP/IP. It is a connection-oriented protocol that per- forms several functions including: providing for reliable transmission of data by furnishing end-to-end error detection and correction; guaranteeing that data are transferred across a network accurately and in the proper sequence; retransmit- ting any data not received by the destination node; and guaranteeing against data duplication between sending and receiving nodes. Application protocols using TCP include Telnet, File Transfer Protocol (FTP), Simple Mail Transport Protocol (SMTP), and Post Office Protocol (POP).

�9 UDP. This is a connectionless protocol providing an unreliable datagram service. It does not furnish any end-to-end error detection or correction, and it does not retransmit any data it did not receive. UDP requires very little overhead since it does not provide any protection against datagram loss or duplication. Application protocols based on UDP include the Trivial File Transfer Protocol (TFI'P), Net- work File System (NFS), the Simple Network Management Protocol (SNMP), the Bootstrap Protocol (BOOTP), and Domain Name Service (DNS).

Page 22: Networking Explained || Network Topologies, Architectures, and the OSI Model

52 Networking Explained, Second Edition

Application (4) �9 Similar to OSI application layer. �9 Serves as communication interface by providing specific application services. �9 Examples include e-mail, virtual terminal, file transfer, WWW.

Transport (3) �9 Defined by two protocols:

User Datagram Protocol (UDP) �9 Is a connectionless protocol. �9 Provides unreliable datagram service (no end-to-end error detection or

correction). �9 Does not retransmit any unreceived data. �9 Requires little overhead. �9 Application protocols include Trivial File Transfer Protocol (TFTP), NFS,

Simple Network Management Protocol (SNMP), Bootstrap Protocol (BOOTP), and Domain Name Service (DNS).

Transmission Control Protocol (TCP) �9 This is the TCP of TCP/IP. �9 Is a connection-oriented protocol. �9 Provides reliable data transmission via end-to-end error detection and

correction. �9 Guarantees data are transferred across a network accurately and in proper

order. �9 Retransmits any data not received by destination node. �9 Guarantees against data duplication between sending and receiving nodes. �9 Application protocols include Telnet, FTP, SMTP, and POP.

Internet (2) �9 Heart and soul is Internet Protocol (IP)mthe IP of TCP/IP. �9 Transfers user messages from source host to destination host. �9 Is a connectionless datagram service. �9 Route selection is based on some metric. �9 Uses Internet or IP addresses as a road map to locate a host within the

internet. �9 Relies on touters or switches (dedicated nodes that connect two or more

dissimilar networks). �9 Integral part is Internet Control Message Protocol (ICMP), which uses an IP

datagram to carry messages about state of communications environment.

Network Interface (1) �9 Connects a host to the local network hardware. �9 Makes a connection to the physical medium. �9 Uses a specific protocol for accessing the medium. �9 Places data into frames. �9 Effectively performs all functions of the first two layers of the OSI model.

I IIIII

FIGURE 2.15 Summary of the TCP/IP layers and functions.

Page 23: Networking Explained || Network Topologies, Architectures, and the OSI Model

Chapter 2: Network Topologies, Architectures, and the OSI Model 53

OSI Layers Included Protocols TCP/IP Layers

Application

Presentation

Session

SNMP TFTP NFS DNS

BOOTP

Transport UDP

FTP Telnet Finger SMTP POP

TCP

Network ~ IP

Data Link Network Interface Cards

Physical Transmission Media

Application

Host-to-Host Transport

Internet

Network Interface

FIGURE 2.16 Another comparison of the OSI and TCP/IP layers. Source: Adapted from Miller, 1992.

The TCP/IP Internet layer (also called the network layer) transfers user messages from a source host to a destination host. The heart and soul of this layer is the Internet Protocol, which is the IP of TCP/IP. IP is a connectionless datagram service responsible for routing packets between nodes. In short, IP receives data bits from the lower layer, assembles the bits into packets (IP datagrams), and selects the "best" route based on some metric. (A metric is a description of the "cost" of a route used by routing hardware and software to select the best possible route.)

The TCP/IP network interface layer connects a host to the local network hardware. Its functions include making a connection to the physical medium, using a specific protocol for accessing the medium, and segmenting data into frames. It effectively performs all of the functions of the first two layers of the OSI model.

51. What is OSrs role given that the Internet is based on TCP/IP?

At one point during the early 1990s, it was believed by many that the OSI protocols were going to become "the" network standard for everyone. Even the U.S. Government got into the act by establishing GOSIP (Government OSI Profile), which mandated all government organizations purchase OSI-compliant networking products beginning in 1992. In 1995, however, GOSIP was modified to include TCP/IP as an acceptable protocol suite for GOSIP compliance. Today, OSI protocols are in use, but their presence pales in comparison to that of their TCP/IP counterparts. Nonetheless, the OSI model has had a lasting impact on networks, including TCP/IP. The model continues to provide a detailed standard for describing a network. It is from this perspective that the network design com- munity continues to regard the OSI model as a theoretical framework for the development of networks and their architecture.

Page 24: Networking Explained || Network Topologies, Architectures, and the OSI Model

54 Networking Explained, Second Edition

52. if TCP/IP is "in" and predates OSl, why bother studying OSI? Recall from Chapter 1 our discussion on protocols. We stated that protocols are neces-

sary to ensure that communications are understood. Analogous to this, the OSI model pro- vides us with a common communication "protocol." It enables us to discuss network principles and concepts with other networking professionals and be understood. For exam- ple, if a vendor claims that its latest switch operates at layer 4 or if a co-worker states that network errors are occurring at layer 2, you will understand what each person is talking about because they are referencing the OSI model.

END-OF-CHAPTER COMMENTARY On this note we conclude our discussion of network topologies, architectures, and the OSI and TCP/IP models. We will expand upon this material in subsequent chapters. For exam- ple, in Chapter 3, we give a detailed presentation about the Intemet and TCP/IP protocols, including protocols that operate at layers 3 and higher. We also expand the concept of the OSI layers in later chapters. Chapter 4 is dedicated to the physical layer; Chapter 5 pre- sents a discussion on the data link layer; network hardware components that operate at either layer 1 or 2 are presented in Chapter 6; and Chapter 7 addresses concepts relating to the network layer. Other chapters also expand on specific layer 2 and layer 3 protocols.