A local area network is simply a collection of computers that are connected by means of a cable or some other medium and that can communicate with each other in a useful manner. An operating system like Windows 98 contains all of the software required to support this communication. All you have to do is supply the physical connection between the computers and install and configure the networking software components on each machine.

While the intention of this book is to approach the subject of PC networking from a practical standpoint, it is not designed to be a static list of instructions for building a LAN. Designing, constructing, and administering a network also requires judicious amounts of common sense and practical reasoning, and not just an operations manual. Therefore, a small amount of networking theory is necessary so that you can understand how Windows 98 systems interact on a network and how the computers package data and transmit it over a LAN. You will need this information when it comes time to organize your resource sharing strategy and troubleshoot problems on your network.

All communications on a PC network are based on the concept of clients and servers. A client is a hardware or software element that requests data in some form and a server is the element that fulfills that request. However, even though all networks have clients and servers, not all networks are classified as client/server networks. On client/server networks (such as those using Novell NetWare), individual computers are dedicated to client and server roles. A NetWare server, for example, runs a proprietary operating system that is wholly dedicated to server functions. You cannot run a DOS or Windows program on this machine or use it for standalone computing tasks. Typically, the server is a larger and faster machine than the network workstations, because it must service dozens or hundreds of users.

The clients on a NetWare network are standard PCs running operating systems like Windows and DOS, but with the addition of client software that makes it possible for the computers to communicate with NetWare servers. Thus, on a client/server network, the client systems can communicate with the servers, but they cannot communicate with each other. For example, if you want to give a file to a user on another PC, you must copy it to a server drive and the other user must retrieve it from there. Both of you have access to the server drive (as shown in Figure 2-1), but neither of you can copy the file directly to the other user's PC.

The other type of network in general use today is called a peer-to-peer network. In this arrangement, every computer on the network runs an operating system that is capable of performing both client and server functions. Windows 98, 95, NT, and 2000 are all capable of functioning as peers in a network, as is Windows for Workgroups (but not Windows 3.1). There is no dedicated server operating system on a peer-to-peer network. You can run a standard DOS or Windows application on any machine, regardless of its networking capability. When a computer on a peer-to-peer network is accessing the resources of another machine, it is functioning as a client; when it shares its own resources with other computers on the network, it is functioning as a server. Thus, while a functional diagram of a client/server network shows the clients as the spokes of a wheel with the server in the center, a peer-to-peer network diagram has the appearance of a mesh in which each of the computers has equal access to every other computer, as shown in Figure 2-2.

The client and server functions on a peer-to-peer network, therefore, are roles that you can assign to various machines and modify at will. You can use a client/server operational model on your Windows 98 network by designating one computer as a server, on which you will store all the data shared by the network's users, but there is nothing inherently different about the computer's hardware or its operating system configuration. Even if you do choose to organize your network in this way, there is nothing stopping you from sharing the resources of the network's other systems as well, or from using the designated server to access other shared resources.

The result is that with a peer-to-peer network, you have complete freedom to assign your computers client and server roles as you see fit. For example, you can designate one computer as the file server where all of the shared network data will be stored, and a different machine as the print server, where print jobs will be queued while they are waiting to be processed by the printer. To do this on a NetWare network would require two dedicated servers, but on a Windows 98 peer network, you can assign these roles to any machine and even change them at a later time without having to install a different operating system.

In order for any two entities to communicate with each other, they must share a common language. In computer networking, these languages are called protocols. For example, when you type in the URL for a web page in your browser, the browser uses a protocol called HTTP (the HyperText Transfer Protocol) to transmit a request to the appropriate server. Since the server uses the same protocol, it understands the request and transmits a reply containing the web page. This type of request and response interchange occurs constantly on a computer network, and at very high speeds.

Unfortunately, networking is not limited to the use of a single protocol. There are literally dozens of protocols used by networked computers, many of which operate at the same time during a single request/response transaction. This is because network communications occur at many levels and for many different purposes, all running simultaneously. The HTTP protocol, for example, functions at the top layer of a stack of protocols, all of which are involved in the communications process. The single web page request you issue at your LAN workstation also uses several other protocols in order to get the message to its destination, including TCP (the Transmission Control Protocol), IP (the Internet Protocol), and Ethernet.

The computers use these multiple protocols by encapsulating the original request in a data envelope (called a header or a frame) created by each successive protocol, resulting in a unit called a packet that gets transmitted over the network. Thus, on a networked system, the original HTTP request containing the URL is carried within a TCP frame, which is in turn carried within an IP frame, which is in turn carried within an Ethernet frame. The packet is then transmitted to the destination system, which reads each successive frame and passes the request to its ultimate destination: the web server software running on the machine.

To illustrate the levels at which these network communications occur, a group called the International Organization for Standardization (paradoxically abbreviated as the ISO) developed a theoretical structure they call the Open Systems Interconnect (or OSI) Reference Model. This model defines the network communications process as consisting of the seven layers shown in Figure 2-3. Requests for access to network services originate at the top of the model at the application layer and are passed down through the layers until they reach the physical medium of the network itself. Although it can take other forms, such as radio transmissions, the network medium is usually a cable connecting the computers. Every computer on the network must use the same protocols at each layer of the model in order to communicate.

Thus, using the OSI model as a guide, you can see that the HTTP protocol functions at the application layer, TCP at the transport layer, IP at the network layer, Ethernet at the data link layer, and 10BaseT at the physical layer. Many people that are new to networking become confused when they read or hear about various types of LANs, such as TCP/IP networks and Ethernet networks and 10BaseT networks. This is because it may not be immediately evident that all three of these terms (and quite a few others) can be referring to a single LAN. TCP/IP, Ethernet, and 10BaseT are not different types of networks, rather they are different types of networking protocols. They operate at different layers of the OSI model and perform different functions. Typical networked computers like those on your Windows 98 network can run all three at the same time.

The protocols running at the various layers of the model perform functions that together provide a quality of service that ensures the accurate transmission of data over the network. The protocols are designed so that a function provided in one is not duplicated in another. TCP, for example, is a protocol that provides guaranteed delivery of data by requiring the destination system to acknowledge the packets it receives. Because TCP takes care of this task, there is no need for IP, running at the layer beneath it, to require acknowledgements as well.

The functions provided by the protocols running at the various layers of the OSI model are numerous, so in the interest of simplifying what can quickly become a very complex subject, this book is going to use a condensed version of the model that consists of only three layers. As shown in Figure 2-4, the physical and data link layers are combined, as are the network and transport layers, and the session, presentation, and application layers. The three layers in this revised model roughly correspond to the types of networking modules that you will work with in Windows 98, and will be referenced throughout this book as the physical link, network transport, and upper layers, respectively.

The protocols running at the physical link layer include Ethernet as well as the 10BaseT cabling system and its variants. These protocols define all of the physical properties of the network, as well as how the computer interfaces with the network medium. These properties include the following:

Cable type

Cable installation

Transmission speed

Media access control

Packet addressing

As you plan and construct your network, your attention to the physical link layer will encompass the following procedures:

• Selecting cables, hubs, network interface cards, and other connecting hardware.
• Laying out the locations of your computers and other devices.
• Installing the network cable and inserting the NICs into the computers.
• Installing network adapter drivers in Windows 98.

A small home or office LAN is best served by what is known as a 10BaseT Ethernet network. Ethernet is a networking protocol that has been around since the 1970s, and is used on the vast majority of local area networks installed worldwide. Virtually every manufacturer of networking hardware supports Ethernet, giving you a wide variety of products to choose from. As the most familiar LAN protocol, it is also not difficult to find help while installing and configuring your Ethernet network, either from online resources such as web sites and newsgroups or from live consultants and technicians.

10BaseT describes the type of cable that you will use to connect your computers, as well as a set of cabling guidelines that define the physical limitations of your network, such as how far one computer can be from another. There are alternatives to 10BaseT that you can choose, which will be discussed later, but this basic type of network is inexpensive, easy to install, and more than adequate for the average user's needs.

Other protocols that operate at the physical link layer include Token Ring, Fiber Distributed Data Interface (FDDI), and Asynchronous Transfer Mode (ATM). Any one of these protocols is capable of providing the appropriate services to a small network, but their expense and complexity tends to make them impractical for a network this size.

The protocols that operate at the network transport layer correspond to the networking modules referred to as protocols in Windows 98. The network transport layer protocols supported by Windows 98 are as follows:

TCP/IP

NetBEUI

IPX/SPX-compatible

You will have to install at least one of these protocol modules on all of your network's systems in order for them to communicate. This is another source of confusion, because the modules running at the other layers, like Ethernet and HTTP, are just as much protocols as the TCP/IP, NetBEUI, and IPX/SPX-compatible modules running at the network transport layer. In addition, although Windows 98 treats TCP/IP and IPX/SPX as single entities, they are actually collections of several different protocols, designed to work together to provide a unified service. Thus, adding protocols in Windows 98 parlance refers only to the installation of single modules like TCP/IP, IPX/SPX-compatible, and/or NetBEUI, even though these procedures actually install support for several different protocols. In addition, many other network configuration procedures (such as installing a network card) involve the addition of what are technically known as protocols.

On a small network, the protocols functioning at the network transport layer are responsible primarily for seeing to it that data is transmitted to its destination quickly and accurately. This typically involves the transmission of additional packets that do not contain user data, but serve instead to control the communication between computers. For example, when a web browser requests a URL from a web server, the two computers perform a complex series of data exchanges using the TCP protocol, as illustrated in Figure 2-5.

The TCP communication process begins with the client system sending packets that query the server to determine if it is functioning properly and ready to receive data. When the server responds in the affirmative, it requests the establishment of a connection in the other direction. When the client acknowledges this request, the two machines are said to have established a bi-directional connection. It is for this reason that TCP is called a connection-oriented protocol. Only after the connection is established does the client transmit its URL request. When the server responds by sending the requested web page, the client periodically acknowledges the received packets by sending additional control messages. After the client has received all of the requested data, the two computers again exchange messages that break down the two connections.

Thus, a transaction of this type can involve the transmission of a significant amount of data that is not actually utilized by the applications on the two systems. As an illustration of just how much control traffic is involved in network communication, consider the fact that the entire transaction described here occurs during the transmission of every file over the network. When you access a web page that consists of text and graphics, the server supplies the graphics as separate image files, each of which must be transmitted using its own separate TCP connection. Thus, your system may run through the entire connection establishment, data acknowledgement, and connection termination process a dozen or more times for every web page displayed in your browser.

Not every exchange of information on a network requires this much overhead. The sister protocol to TCP, for example, is called the User Datagram Protocol (UDP). UDP is a connectionless protocol that computers use for rapid query and response transactions that do not require the transmission of a lot of data. However, depending on the protocols you use at the network transport layer, a large percentage of the traffic sent over the cable can consist of control traffic that contains no user data.

The only tasks associated with the network transport layer in your network installation process are the selection and installation of the protocol modules you will run on your Windows 98 systems. This selection is based on the types of services you want to provide to your users. While Windows 98 does make it possible to fine tune certain operational parameters of these protocols, these adjustments are not necessary for efficient operation of a network.

The upper layer, which incorporates top three layers of the official OSI reference model, is the most difficult of the three layers in the simplified model to describe because its functions are so diverse and are scattered about several different types of software components. The upper layer includes the applications that directly access network resources, such as web and FTP clients and servers, as well as the infrastructure that makes it possible for these applications to utilize the protocols at the network transport and physical link layers.

Some types of network applications have protocols associated with them, that are dedicated to specific types of tasks. Some of these protocols are as follows:

• Hypertext Transfer Protocol (HTTP) - Used by world wide web servers and browsers to transmit URL requests and web page contents.
• File Transfer Protocol (FTP) - Used to transfer large files across TCP/IP networks.
• Simple Mail Transfer Protocol (SMTP) - Used to transmit e-mail messages across TCP/IP networks.
• Simple Network Management Protocol (SNMP) - Used to transmit status information about networked devices to a centralized management console.

There are many other protocols that operate at the upper layer, some of which define specific application types (like FTP) and others that work behind the scenes to provide network services (like SMTP). You will encounter some of these when you install various types of network applications on your workstations.

In addition to protocols, there are other elements operating at the upper layer that make network communications possible. For example, when you add the Client for Microsoft Networks module to a Windows 98 system, you are installing what is known as a redirector. A redirector is the component that makes it possible for applications that are not themselves network aware to access network resources. When you open a file in a word processing application, for example, the redirector determines whether the requested file is located on a local or network drive and sends the request either to the computer's storage subsystem or the networking protocol stack.

Windows 98 also uses application programming interfaces (APIs) at the upper layer to provide applications with access to network resources. For example, the Windows Sockets interface (often referred to as Winsock) enables programs to access resources on a TCP/IP network like the Internet. All web browsers, FTP clients, and news readers are Winsock applications, because they rely on this unified programming interface. Without Winsock, each application would require its own individual network client components, which would cause all sorts of compatibility problems and inflate the size of the programs considerably.

NetBIOS is another programming interface used by Windows 98, that provides the name space the operating system uses to identify the computers on the network. When you specify a computer name during the Windows 98 installation, you are really assigning the system a NetBIOS name. Every computer on the network must have a unique name, or the operating system generates an error message during the network logon process.

Unlike the network transport and physical link layers, there are few choices to be made at the upper layer. You can elect to use Ethernet at the physical link layer or choose another protocol such as Token Ring or FDDI. You can also decide to use 10BaseT cable, one of the new Fast Ethernet variants, or an older technology like coaxial cable. At the network transport layer, you must decide whether to use TCP/IP, IPX/SPX, NetBEUI or a combination of these protocols. At the upper layer, however, the protocols and interfaces that Windows 98 uses are mostly predetermined. A web server or browser must use HTTP for its protocol and a Winsock module to access the TCP/IP network, for example.

The combined functions of the upper, network transport, and physical link layers can be said to form a networking stack that is illustrated by the simplified reference model. The top of the structure represents the application running on the computer and the bottom is the cable connecting the computer to the network. When the application requests access to a network resource, the system passes the request down through the layers to the network medium itself (see Figure 2-6).

The protocols operating at the various layers encapsulate the data they receive from the layer above in order to package it for its journey across the network. This packaging consists of data added to the message that is intended for the same protocol running on the destination computer. Just as a postal worker may stamp an envelope with a message like "First Class Mail" that is intended for other mail handlers during its journey, data packets include additional data (called headers or frames) that other computers use to get the message where it has to go.

The layers of the networking stack must form an unbroken sequence from the top of the model to the bottom. If a required protocol at any of the layers fails to function, whether it be due to a missing program file, a malfunctioning network interface card, or a broken cable, then network communications cease.

Once the message reaches the network cable and is transmitted to the destination computer, the entire process is duplicated in reverse (see Figure 2-7). The receiving system passes the data packet up through the layers, with each protocol reading the header or frame intended for it. The control data in the headers and frames tells the system where the packet should go next.

For example, the Ethernet frame contains a code that specifies which network transport layer protocol the system should use to process each arriving packet. The network transport protocol header in turn specifies the upper layer process that should receive the message. Thus, the receiving system must also have an unbroken networking stack in order for a packet to reach the application that is its ultimate destination. If one of the protocols is not functioning properly, or if the receiving system does not have the same protocols installed as the sender, no network communication will occur.

Note, however, that this does not mean that the communicating computers must be running the same operating system or application, just the same protocols. A Windows 98 system running TCP/IP at the network transport layer and an HTTP-based web browser at the upper layer can communicate with a UNIX server, as long as it is also running TCP/IP and an HTTP-based web server. The protocols operating at the physical link layer do not have to be the same, because the data packets typically pass through intermediate systems called routers or switches that are designed to translate between the physical link layer protocols.

A basic Ethernet network requires three types of hardware devices in addition to the PCs:

• A network interface card (NIC) for each computer.
• A hub that will act as the central connector for all of the computers.
• Cables to connect each computer to the hub.

A network interface card functions as a computer's conduit to the network medium (usually a cable). The combination of the NIC and the network adapter driver installed on the machine combine to form the physical link layer of the computer's networking stack. The network cable connects all of the computers together using a connecting device called a hub (or sometimes a concentrator). A hub is simply a cable nexus to which all of the computers on the network are connected. A hub for a small network usually takes the form of a box with a number of RJ-45 jacks (called ports) for the network cables and an LED for each port, as shown in Figure 2-8.

Hubs for large networks are usually rack-mounted affairs with 12, 16, or more ports and other additional features, but the type of hub you will need for a small network is usually no larger than a small paperback book and has 4, 8, or 12 ports. The hub requires a power source, because it boosts the strength of the electrical signals passing through it. The simple hubs used on small networks simply take the signals entering through any of their ports, amplify them, and transmit them back out through all of the other ports. A device that amplifies the signals on a LAN is also called a repeater., and this type of hub is sometimes called a repeating hub. Higher end devices called intelligent, smart, or switching hubs can read the contents of an incoming data packet and transmit it out only through the port connected to the destination system. This sort of intelligence is not needed on a small LAN.

Every computer connects to the hub using a network cable, forming what is known as a star topology (see Figure 2-9). The term topology describes the physical layout of the network. Other network types use bus or ring topologies or hybrids combining two or more topologies.

Figure 2-9: The topology of a network specifies how the computers are physically connected to each other.

The cable that most 10BaseT networks use is called unshielded twisted pair (UTP), which consists of four pairs of copper wires within a plastic covering, with each pair twisted at regular intervals to minimize the effects of interference from each other and from outside sources. The cables use RJ-45 connectors (see Figure 2-10), which have eight pins and are similar in appearance to the standard 4-pin RJ-11 connectors used on telephone cables. UTP cables are available preassembled (that is, with the connectors attached) or you can buy the cable in bulk and attach the connectors yourself.

All of the hardware required to network your PCs, with the exception of the cable, is designed specifically for use on Ethernet networks. If you were going to install a different physical link layer protocol, such as Token Ring, you would have to purchase NICs and hubs designed specifically for that type of network.

Ethernet has been around for over 20 years, and the standards defining the protocol have undergone several revisions in that time. By far, the most common network type in use today is 10BaseT Ethernet, which runs at 10 Mbps over unshielded twisted pair cable. However, 10BaseT falls near the center of the Ethernet development time line, as shown in Figure 2-11.

Earlier versions of the Ethernet standard called for coaxial cable in one of two forms: either thick Ethernet (also called 10Base5) or thin Ethernet (also called thinnet, cheapernet, or 10Base2). Newer versions of Ethernet, called Fast Ethernet (100BaseT) and Gigabit Ethernet (1000BaseT), use the same basic cable type as 10BaseT, but run at higher speeds (100 and 1000 Mbps, respectively).

As mentioned earlier, for average networking tasks, you can't go wrong with a 10BaseT network, but if you have special considerations, the following sections examine the practical alternatives to 10BaseT that you can select instead.

Thick Ethernet hardware is no longer readily available, but it is possible to obtain thin Ethernet NICs and cables. The primary advantage of thin Ethernet is that it does not use hubs, which does cut down on expenses. Thin Ethernet networks use a bus topology, in which each computer is connected directly to the next one. The NIC in each computer on a thin Ethernet network has a T-connector attached to it, that enables you to attach the two systems on either side using coaxial cables about ¼ inch thick with BNC (Bayonet-Neill-Concelman) connectors (see Figure 2-12). The two ends of the bus must have terminator plugs on them containing resistors that cancel out signals so that they do not reflect back across the network.

Thin Ethernet has largely been phased out in favor of 10BaseT in recent years, but it still remains a viable alternative for your Windows 98 network. However, there are several important drawbacks to thin Ethernet that you should consider before using it to cable your network:

• Pre-made cables (with connectors attached) are much less readily available than those for 10BaseT. Making your own cables by purchasing a spool of bulk cable and attaching the connectors yourself is possible, but it requires a special crimping tool and a reasonable amount of practice.
• Thin Ethernet cable connections (often because they're home made) can be notoriously quirky. Even a light pull on a cable can cause an intermittent break in communications that is very difficult to track down and diagnose.
• Coaxial cable is more difficult to conceal, because every computer (except those at the ends of the bus) must have two ¼ inch cables connected to it.
• Networks that use a bus topology are subject to the Christmas-light effect, in which a single break in the cable interrupts communications to all of the computers on the far side of the break. Because of the use of a hub, a cable break on a star network like 10BaseT only affects one computer.
• The speed of a Thin Ethernet network is limited to 10 Mbps, since the cable is physically incapable of handling faster communications. With careful planning and purchasing, you can install a 10BaseT network today and upgrade it to 100 or 1000 Mbps at a later time.

There are virtually no NICs left on the market designed to work only with thin Ethernet, but there are many combination devices available that support coaxial cable connections for thin Ethernet in addition to twisted pair (and sometimes even the AUI connections used for thick Ethernet). However, these combination devices tend to be more expensive than twisted pair-only NICs, so there is little money to be saved there.

Thus, in terms of manageability, upgradability, and ease of installation, thin Ethernet is usually not worth the little bit of money you'll save by not buying a hub. If you happen to come across someone who is upgrading a network from thin Ethernet to twisted pair, you may be able to get their old NICs and cables for little or nothing, and in this case it might be worth putting together a coaxial network, but buying new thin Ethernet equipment is not recommended.

On the other side of the time line are the new Ethernet variants running at significantly higher speeds than traditional 10BaseT. Fast Ethernet, sometimes referred to as 100BaseT, runs at 100 Mbps, providing sufficient bandwidth for applications that transfer large amounts of data across the network, such as streaming video and graphics-intensive programs. The term 100BaseT, however, can refer to any one of three possible network medium configurations depending on the type of cable the network uses. The three types of Fast Ethernet network are as follows:

• 100BaseTX - Requires two wire pairs in a Category 5 UTP cable.
• 100BaseT4 - Requires four wire pairs in a Category 3, 4, or 5 UTP cable.
• 100BaseFX - Requires the use of fiber optic cable.

If you decide to install a Fast Ethernet network, it will probably be 100BaseTX, because virtually all of the network cables sold today are Category 5. 100BaseT4 is designed for use on networks with an existing Category 3 cable installation. This option makes it possible to upgrade an older 10BaseT network to Fast Ethernet without pulling new cable.

Fiber optic is a completely different type of network medium that uses pulses of laser light to transmit data instead of electrical signals. You can't just buy a fiber optic cable from the computer store and plug it into your PC. Installing and maintaining fiber optic cable is much more complex and expensive than working with standard copper cable. Since 100BaseFX runs at 100 Mbps, just like the other types of Fast Ethernet, the main advantages that fiber optic provides over the other two is that it can span much greater distances than any copper cable (up to 412 meters in the case of 100BaseFX) and that it is immune from the electromagnetic interference that can affect copper cable networks.

Apart from the cable requirements, Fast Ethernet requires network interface cards and hubs that support the new standard. You must also consider the capabilities of the PCs that you will be using when deciding whether or not to opt for Fast Ethernet. If your PCs are relatively new and you are building a new network from scratch, the difference in price between 10 Mbps and Fast Ethernet equipment is minimal, and it may be worthwhile to go with Fast Ethernet from the outset. Since Fast Ethernet uses the same media access control mechanism and frame format as regular Ethernet, the troubleshooting and support procedures are the same, as are the basic cabling guidelines.

Much of the Fast Ethernet equipment on the market is dual-purpose, running at either 10 or 100 Mbps. For this reason, you can easily purchase hardware that enables you to run your network at 10 Mbps now and upgrade it to Fast Ethernet later. For more information on purchasing components with upgradability in mind, see Chapter 4, Purchasing Network Components.

The typical file and printer sharing tasks performed on local area networks do not require the speed of Fast Ethernet, but PC networking technology advances at a rapid pace and the next killer application may well benefit from the additional speed. If you can afford the extra cost, it is a good idea to purchase equipment that is capable of running at 100 Mbps, even you only intend to use the slower speed right now.

A good part of practical networking involves making decisions, such as which products to buy, which operating system modules and applications to install, and which of several methods should you use to accomplish a particular task. The discussion of Ethernet in this chapter has illustrated just a few of the decisions that you, as a network administrator, must make as you construct your Windows 98 network. The following chapters introduce more decisions you must make as you plan your network, select products, and implement policies.