Network Connection Technologies

As you know, a LAN consists of a group of computers connected together using some sort of electrical medium. You can choose from several different electrical media. They differ in the way they format and electrically represent the data sent between computers.

Network devices have to use some standardized way of organizing the data signals they transmit between computers. You might have heard of some of these already:

• Ethernet was developed by Xerox, Intel, and Digital Equipment Corporation. Ethernet has grown so popular and common that you hardly need to use the word anymore: Most networks are Ethernet networks.

• Asynchronous Transfer Mode, or ATM, is a networking technology widely used in the telecommunications and Internet industries for very high-speed backbone networks. Backbone is a term for an ultra-fast connection between the separate parts or sites of a large network. For example, it might refer to the set of links between major network sites of a corporation, the national network of a telephone carrier, or the high-speed Internet links between major ISPs. ATM is often used behind-the-scenes by Internet service providers to route data to DSL (Digital Subscriber Line) modems. However, ATM is only rarely used to connect directly to individual user's computers, except for some very specialized graphics workstations and other ultra-high-tech situations.

Other technologies are waning in popularity and you won't be hearing of them again, so don't feel any need to memorize:

• ARCnet, AppleTalk, and StarLan were early network technologies but are hardly used now because they are so much slower than modern technologies. AppleTalk is still occasionally used to connect older Macintoshes to printers. ARCNet is still used in some industrial settings, for example, to control factory equipment.

• Token ring was developed by IBM and is still used in businesses that are "blue" to the core, but nobody else in his right mind would install it now because it's slower and much more expensive than Ethernet.

If you're constructing your own network, you'll likely use Ethernet in one form or another. The choice you'll have to make is which kind of physical medium to use.

Physical Media

The signals transmitted across a LAN are generated and interpreted by electronics in each computer. Some computers have built-in network interfaces; otherwise, each computer in a LAN needs a network interface card, or NIC. I may also refer to them as network cards.

These electrical signals have to be carried from computer to computer somehow. The original design for Ethernet used a very expensive 1/2-inch thick cable that could carry a 10Mbps Ethernet signal up to 500 meters. (It was named 10BASE5 for reasons that only make sense to an engineer.)

Today's network interface cards are designed to use one of several inexpensive varieties of network cabling, or use radio waves to avoid the need for wiring altogether. In the following sections, I'll list the various types of media you're likely to encounter.

Thin Ethernet, or ThinNet

Thin Ethernet used a coaxial cable similar to television cable to connect each computerI say "used" because it's an old, cumbersome technology that has virtually disappeared. It was also called 10BASE2 Ethernet; the 10 indicated that the network ran at 10Mbps, and the 2 indicated that it had a maximum wiring length of 200 meters or 660 feet. Thin Ethernet cables ended in distinctive twist-on connectors called bayonet connectors, or BNC.

Thin Ethernet wiring ran from computer to computer in a daisy-chain fashion called a bus network, as shown in Figure 15.2.

Figure 15.2. Thin Ethernet network cabling ran from computer to computer in a single unbroken line called a bus.

Some coaxial cable may still be around, but it's been almost completely supplanted by the much faster, much less expensive, and much easier to use twisted-pair system.

Unshielded Twisted-Pair (UTP)

Unshielded Twisted-Pair, or UTP, has become the most common network carrier, and is so called because like-colored pairs of wires inside the cable are gently twisted together for better immunity to electrical interference from fluorescent lights, radio signals, and so on. This inexpensive type of cable is also used for telephone connections, although the network variety is of a higher quality and is certified for its capability to carry high data rates. UTP cables are terminated with eight-wire RJ45 connectors, which are wider versions of the ubiquitous modular telephone connectors.

UTP cable quality is categorized by the highest data rate it's been designed and certified to carry reliably. The most common cable types are shown in Table 15.1.

Table 15.1. UTP Cable Categories

Designation	Highest Data Rate	Application
CAT-1	Less than 1Mbps	Telephone (voice)
CAT-2	4Mbps	IBM Token Ring
CAT-3	16Mbps	10Mbps Ethernet (10BASE-T)
CAT-4	20Mbps	16Mbps Token Ring
CAT-5	100Mbps	100Mbps Ethernet (100BASE-T), ATM, others
CAT-5E or -5X	250Mbps	Gigabit Ethernet[*] (1000BASE-T)
CAT-6	250Mbps	Gigabit Ethernet[*]

^[*] Gigabit Ethernet uses four pairs of wire each carrying 250Mbps, providing an aggregate speed of 1000Mbps

The thing to remember here is that you can't use just any old wiring you find in your walls to carry a network signal: You have to look for the appropriate "CAT-something" designation, which will be printed on the cable jacket every foot or so.

UTP cabling can carry token ring signals but is most commonly used for Ethernet networking. UTP Ethernet devices are connected to a central device called a hub in what is called a star network, as shown in Figure 15.3. Star networks are reliable: If any cable in a bus network broke, the whole network failed. If a cable in a star network fails, only the computer connected by that cable goes offline.

Figure 15.3. Unshielded twisted-pair network with a hub.

You can buy three varieties of UTP-based Ethernet hardware, denoted 10BASE-T, 10/100BASE-T, and 1000BASE-T in order of increasing speed. I'll discuss the 10BASE-T variety first.

10BASE-T

If you've been paying attention, you might guess that the 10BASE part means 10Mbps, but T? The T stands for twisted pair, and you just have to know that the maximum permitted cable length is 100 meters, or 330 feet.

This is usually an ample distance in a home or small office environment, but it limits 10BASE-T's usefulness in a large building or campus LAN. Hubs can solve this problem by serving to connect several close-by computers. The hubs can then be connected to each other with fiber-optic cable, which forms a "backbone" connecting groups of computers, as shown in Figure 15.4.

Figure 15.4. In larger LANs, hubs are connected together to span larger distances. Hubs can be connected using UTP or fiber-optic cabling.

Half Versus Full Duplex

You might see network hardware advertised as "full duplex," or labeled 10BASE-Tx or 100BASE-Tx. If you're curious, here's what this "full" business is about.

Because thin Ethernet cable had only one wire, like our tin can network, only one device at a time could "speak." In networking parlance, this was called a half-duplex network. 10BASE-T was usually half-duplex, too. Some manufacturers began making 10BASE-Tx cards that could operate in a full-duplex manner, where two network cards can "speak" to each other simultaneously. All of the cards and hubs in the network have to be designed to be capable of it. Today, virtually all Ethernet hardware sold is full-duplex capable, even if it's labeled T rather than Tx.

Operating this way can boost communication rates somewhat, but it does not necessarily double them. It turns out that most computer-to-computer conversations are just like the ones you dread being trapped in at parties: One person does most of the talking, while you just nod and say "uh huh" every so often. In a full-duplex network, the "uh huhs" don't have to interrupt the nonstop talker, so the conversation goes a little bit fasterbut not much.

100BASE-T, Fast Ethernet

Fast Ethernet is a 100Mbps version of Ethernet over UTP cable. It is also called 100BASE-T or 100BASE-Tx. Most current hardware can actually work at either speed, and is labeled 10/100BASE-T or -Tx. (The x stands for full-duplex, which is standard with 100BASE-T networking hardware, with or without the x.)

This hardware is 10 times faster than 10BASE-T hardware. The CAT-5 cable and connectors required to carry this high-speed signal are a tiny bit more expensive than CAT-3 and require more care in their installation, but the cost has fallen so much in recent years that it's really no longer a consideration. 100BASE-T hubs and network cards used to be more expensive as well, but again, they're manufactured in such enormous volumes now that the price differential has disappeared. In fact, most new computers have a 10/100BASE-T adapter built right into the motherboard. The wiring is cheap, so, for a home or small office, Ethernet networking is virtually free.

TIP

It doesn't make sense to buy new 10BASE-T parts now. For new networks, or if you're adding on to an existing 100Mbps version of Ethernet over UTP cable. It is also 100Mbps version of Ethernet over UTP cable. It is also 10BASE-T network, 10/100 equipment is the stuff to get.

1000BASE-T, Gigabit Ethernet

Gigabit Ethernet, 100Mbps version of Ethernet over UTP cable. It is also as you might guess, sends data at 1000Mbps. It's several times more expensive than 100Mbps Ethernet, but the tenfold increase in speed is worth it if your server is trying to feed files to several hundred people at once, or if you're involved in medical imaging, digital video editing, or other applications that involve transferring huge amounts of data. It's also used for the backbones of large networks and for fast server-to-server and server-to-switch connections.

Gigabit networking is overkill in the home and small office environment, as most desktop computers can't transfer data to or from their hard disks fast enough to take advantage of such a fast network. Still, some higher-end workstations, like Mac G5s and most server-class machines, now come with a 10/100/1000BASE-T adapter built-in. This adapter will work at any of these three speeds, depending on the abilities of the hub or switch to which it's connected. The price of Gigabit switches and hubs has fallen to less than $15 per port. For example, a 5-port switch costs about $60. So, if your computers came with Gigabit adapters, have ultra-fast hard drives or RAID arrays, and you expect to be moving lots and lots of data around, it might be worth your while to pay just a bit more for the switch and CAT-6 cabling. Gigabit Ethernet requires CAT-5E or CAT-6 quality cable and connectors throughout, with all four wire pairs connected.

802.11, Wireless Ethernet

It always seemed silly to me to have a portable computer tied down by network and power wires. Now, it doesn't have to be. Ethernet-over-the-proverbial-etherthat is, wireless networkinghas become amazingly inexpensive and ubiquitous. Using wireless network adapters, you can connect computers in a small area (such as a home or office) via radio, as illustrated in Figure 15.5. With modern equipment, the data rate can reach a respectable 54Mbps.

Figure 15.5. Wireless Access Point connecting computers to a standard twisted-pair network.

Wireless access is especially handy for users of laptop computers, Palm Pilots, and other mobile users who visit several offices in the course of a day. A device called an access point can be installed at each location to make the connection between wireless devices and a standard wired network or the Internet. Then, to quote Buckaroo Banzai, "wherever you go, there you are."

Hot spotssites with Wireless access to the Internetare springing up everywhere. In fact, a certain big coffee chain from Seattle is rolling this out nationwidethey'll connect you to the Internet for a small hourly fee while you sip a latte! (Your humble authors would never set foot in one of these places, of course, preferring to patronize locally owned establishments and the original Peet's Coffee & Tea. But I digress.)

NOTE

The wireless network manufacturer's organization is called the "Wi-Fi Alliance." Wi-Fi stands for Wireless Fidelity, and in a loose way, "Wi-Fi" is used to refer to Wireless networking.

One thing you have to watch out for is that there are currently three Wireless standards, named 802.11a, 802.11b, and 802.11g. The "standard" part refers to the fact that the technology is governed by an international standards committee, and equipment made by one manufacturer should work correctly with equipment made by another. (This didn't actually hold true a few years ago, but today it largely does.)

However, equipment designed for one standard won't necessarily work with equipment designed for a different standard, as shown in Table 15.2. 802.11a equipment can only communicate with other 802.11a devices. 802.11b and 802.11g equipment can interoperate, but only at the lowest-common-denominator speed.

Table 15.2. Wireless Networking Standards

Wireless Standard	Data Rate[*]	Compatible with
802.11a	up to 54Mbps	802.11a only
802.11b	up to 11Mbps	802.11b at 11Mbps 802.11g at 11Mbps
802.11g	up to 54Mbps	802.11g at 54Mbps 801.11b at 11Mbps

^[*] Some manufacturers have tweaked their wireless devices to let them communicate at twice the standard's maximum speed, but only when connected to equipment by the same manufacturer.

If you're considering wireless, 801.11g (also called "wireless-g") is the stuff to get for home and small office networks. Wireless-a equipment tends to be expensive. The price differential between -b and -g is miniscule, but -g can go five times faster and it's compatible with -b adapters.

Given the complexity of the stuff, and knowing that just a few years ago it cost about $400 per computer to go wireless, I think today's prices for wireless gear are insanely low: about $2040 per computer for adapters, and $20100 for an access point, the wireless network's hub. While it isn't quite as fast as 100Mbps Ethernet, wireless is so much easier to install that it's competitive with wired networks even in the home and office.

However, there are two things that you must keep in mind. First, a wireless network is not as reliable as a wired network. In my experience and that of many friends, it simply stops working at random intervals; sometimes once a day, sometimes once a week. It may start working again by itself after a few seconds or minutes or hours, or you may have to restart your computers and wireless router to get it back on the air. In contrast, unless someone trips over a cable and yanks the connector off, a wired Ethernet network should run for years without a single glitch.

Second, unless you take explicit steps to secure it, a wireless network is "open to the public," and it's a trivial matter for random passers-by to browse through your shared files and borrow your Internet connection. Making a wireless network secure takes some effort, and to be frank, it can be difficult and confusing even for networking pros, let alone for the technologically challenged. As a result, many people skip the security step just to get their network working, and end up getting their computers hacked-into. A wired network has neither the setup headaches nor the security risks.

Powerline and Phoneline

Network data can also be transmitted as radio signals through your existing telephone lines or electrical wiring. Meant primarily for home use, powerline networking equipment (called HomePlug by its manufacturers' association) and phoneline networking (called HomePNA) send data at up to 10Mbps. It has the advantage of being very easy to installan adapter plugs into your wall jack or phone jack, and connects to your computer. No other wiring or setup is involved.

I rather soundly derided this equipment in the first two editions of this book, but the technology has improved, and the prices have fallen to the point where it makes perfect sense to use it in the average home.

Optical Fiber

Optical fiber is capable of gigabit (1000Mbps) and higher speeds and can also carry data over runs of several miles, quite a bit farther than standard Ethernet. Optical fiber is not generally run directly to individual computers, but between hubs and routers between buildings, to form the "backbone" of a campus network, as shown in Figure 15.6. Optical fiber cables can carry multiple 10 or 100Mbps Ethernet data signals, as well as more advanced, even higher-speed data formats called Fiber Distributed Date Interface (FDDI) and Asynchronous Transfer Mode (ATM).

Figure 15.6. Optical fiber cable linking two 100BASE-T hubs via a fiber "uplink" port. Cables contain pairs of fibers because each fiber can carry information in only one direction.

In summary, there are several different network technologies involved in any network: data transmission format standards like Ethernet and Token Ring, and electrical wiring standards like 10BASE-T and ThinNet. Networks depend on an agreement to use several specific technologies, each of which relies on another to help it do its job. For example, a file-sharing standard relies on a network protocol, which depends on a data transmission format, which requires a wiring standard.

In fact, there's even a standardized way of talking about the way these standards interrelate. In case you haven't guessed already, engineers like nothing more than forming committees to create standards.

The OSI Model

If you've read about networks in any other computer book, you've probably seen a diagram similar to the one in Figure 15.7, the OSI Standard Network Model. The International Organization for Standardization (ISO) and Institute of Electrical and Electronic Engineers (IEEE) developed this modelI think to help computer book authors fill lots of pages trying to explain it. It's in every computer book I've ever seen.

Figure 15.7. The OSI Standard Model for Computer Networksa required figure in every computer book. Networks are built from components, each of which performs a job for a higher-level component in the "stack."

I will spare you the usual long explanation of this diagram because I don't think it's very helpful as an introduction to networking. But I do think it helps illustrate that networks are composed of modular components, conceptually "stacked" one on top of the other, each performing a job for the component above it, using the components below. The parts are interchangeable in that you may often choose one of several available technologies to do the job of a given layer. As long as the job is done correctly, the higher layers don't really care how it's done.

The components in this "stack" communicate with their corresponding components in the other computers on the network. As you go down in these stacks, the layers are less concerned with interpreting the data they handle and more with simply moving it somewhere. The higher level components interpret and communicate with each other to reassure each other that the data they have sent was correctly received, and they rely on the lower levels to actually transport that data from one computer to another.

That's the OSI network model in two paragraphs.

In the real worldat least in the Windows worldthe "stack" of components that make up Windows networking isn't just a concept, it really does exist. Figure 15.8 shows the Windows network model. When you want to access a remote network resource somewhere inside the operating system, the following actions occur:

• A network client composes data messages to communicate these desires to the remote computer, using an agreed-upon file sharing protocol.

• These messages are packaged according to a transport protocol, which specifies how messages are to be broken into manageable pieces, how the pieces are to be addressed to member computers, and how to re-request missing or garbled pieces as they are received.

• The packaged message pieces are called packets and are physically carried by a data link or framing protocol that determines how to arrange the bits of information in each packet for transmission.

• The bits are converted into electrical pulses, radio signals or flashes of light and carried from one computer to another through a physical medium that carries the pulses or flashes to another computer.

• The pulses or flashes are received at the other end; the data work their way up the network components on the other side and are finally delivered to a server component. The server sends a response back through the same path to the client.

Figure 15.8. A practical Windows network model, with actual Windows network components.

The data link level is handled entirely by the hardware in a network interface card (NIC). When you buy a network card, you're buying a data link protocol and the attachment to the physical medium. Because the card is what you'll actually see and have to describe to Windows, from this point on, I'll talk about adapters rather than data link protocols.

Network Clients

A network client is one of the most important top-level parts of Windows networking. The client is responsible for making remote files, folders, and printers available to your computer. To do this, it communicates with a corresponding server component on another computer, whose job it is to deliver file and printer information to client computers. Your Windows XP computer actually has both components built in, because it can both share files and printers and use shared files and printers.

Microsoft provides two network clients with Windows XP Professional: the Client for Microsoft Networks and a Client for Novell NetWare networks. Novell supplies its own version of the NetWare client, downloadable from their Web site, so you actually have a choice of three.

These client components, at the top of their network stacks, communicate with their corresponding top-level server components in other computers to read and write files, queue printer data, read the contents of folders for display in Explorer, and so on. The Client for Microsoft Networks uses the Server Message Block (SMB) and NetBT (NetBIOS over TCP/IP) protocols to speak to other Windows computers, Windows 2000 Servers, and IBM OS/2 LAN Manager Servers. You won't ever encounter SMB or NetBIOS directly in your dealings with Windows Networking; they're part of the client and server software.

The Novell client can communicate with Novell NetWare-based file servers using the built-in NetWare Core Protocol (NCP), or with a Windows XP or Windows 2000 Server network service called File Services for Novell Networks.

NOTE

File Services for Novell and File Services for Macintosh are available only with Windows 200x Server and their more advanced versions. These services allow a Windows Server to share files and printers with Novell workstations and Apple Macintosh computers, respectively.

Each of these client packages uses a transport protocol to carry messages between your workstation and a remote computer's sharing service.

Protocols

As you learned previously, transport protocols define how data is arranged and sent in a coordinated fashion between computers. There are three transport protocols commonly used on Windows-based computers:

• TCP/IP (Transport Control Protocol/Internet Protocol) is the transport protocol that forms the basis of the Internet. TCP/IP is actually a set of many protocols that are used to provide the services that higher-level network components need: resolving computer names into network card and IP addresses, guaranteed transmission, and internetwork routing. The TCP part, or Transmission Control Protocol, is the method an IP-based network uses to guarantee that data is sent end-to-end without errors. I'll go into more detail about TCP/IP in a little bit.

• IPX/SPX (Internetwork Packet Exchange/Sequenced Packet Exchange) was developed by Novell for its NetWare network software. Windows can use IPX/SPX for its file sharing services as well. Like TCP/IP, IPX/SPX is really a set of protocols that provide many services, including name resolution, guaranteed transport, and Internetwork routing.

• NetBEUI (NetBIOS Enhanced User Interface) was developed by IBM for its original IBM PC Network; it provides similar services to TCP/IP and IPX/SPX, except that it doesn't have a mechanism to route data to remote networks. NetBEUI can transport data between computers only on the same physical LAN. NetBEUI was supported by previous versions of Windows, but has been dropped in Windows XP. (Well, almost droppedit's on the Windows XP installation CD in case you have to use it, but it's not easy to find.)

The Client for Microsoft Network can use either TCP/IP or IPX/SPX to send its messages to a file server; all that's required is that both the client computer and server computer have at least one installed protocol in common.

Similarly, Novell's client package communicates with Novell NetWare-based file servers using IPX/SPX, or for recent versions of NetWare, TCP/IP.

The following are some other protocols and acronyms you might run across:

• AppleTalk and its Ethernet-based counterpart LocalTalk are used in Apple Macintosh networking. Windows 2000 Professional provided LocalTalk to facilitate using LAN-connected Apple printers, but it's been dropped from Windows XP.

• DLC (Data Link Control) is an IBM networking protocol, but you won't run into it directly unless you're working on a corporate network with IBM mainframes. In that case, if it's used, your company's network management staff will install and manage it for you. DLC is also used by some network-connected printers. Like AppleTalk, support for DLC has been dropped by Microsoft.

• Point-to-Point Protocol, or PPP, is used to carry Internet Protocol data packets across a dial-up modem connection. This protocol is used to establish almost all modem connections to Internet service providers. PPP is part of the TCP/IP suite and a standard part of Windows's Dial-Up Networking support.

• Point-to-Point Protocol over Ethernet, or PPPoE, is used by some DSL and cable modem Internet service providers to link your computer to the ISP's routing equipment. Its purpose is to limit the number of computers connected to the Internet to just those being actually used. For previous versions of Windows, ISPs provided their own PPPoE software. PPPoE is now built into Windows XP as part of its Broadband Connection support.

• Wired Equivalent Privacy or WEP is an encryption protocol used by wireless networks to protect data from being intercepted by eavesdroppers, and to prevent random passersby from being able to connect to and use your network without permission. In urban areas, it's now common to find that your computer can pick up a half-dozen or more wireless network signals. That means that a half-dozen or more random other people can pick up your wireless signal, and you don't want them poking into your files. WEP helps prevent that. It's unfortunately not completely unbreakablea hacker with a laptop can park him- or herself in front of your house or office for a few hours and eventually be able to get on your network, so another encryption standard was developed, and is becoming more common:

• Wi-Fi Protected Access, or WPA, is an improved encryption scheme for wireless networking. Windows XP Service Pack 2 includes WPA support. Most wireless equipment vendors now support it as well, however, if you have existing equipment, you may have to download upgraded software to get it.

• Universal Plug and Play, or UPnP, lets networked computers and networked devices such as network routers, printers, and household appliances automatically configure themselves to join whatever network they find themselves plugged into. It can, for example, automatically configure your computer to use an Internet connection shared by another computer on the LAN. UPnP also lets these "smart" appliances tell your computer what they do, and can let you configure them from your Windows PC.

• Point-to-Point Tunneling Protocol, or PPTP, is used to create Virtual Private Networks, or VPNs. PPTP takes data destined for a private, remote network, repackages the data for transmission across the Internet, and at the other end unpackages the data to be released into the private, protected network. I'll go into greater detail explaining VPNs in Chapter 18, "Windows Unplugged: Remote and Mobile Networking."

• Layer 2 Tunneling Protocol (L2TP) is another protocol used to create Virtual Private network connections. Windows XP comes with built-in support for both PPTP and L2TC. It is always used along with IPSEC for encryption.

• IPSEC, which stands for Internet Protocol Security, is a protocol that provides very strong encryption of data sent through a TCP/IP network.

• IPV6, for Internet Protocol Version 6, is a new version of TCP/IP. It's sort of TCP/IP on steroids. (Today's ubiquitous TCP/IP protocol is actually version 4. Versions 1 through 3 came and went before the Internet as we know it today existed.) IPV6 is designed primary to get around the most serious limitation of IPV4, which is the limited number of unique IP addressesthe Internet's equivalent of the telephone numberthat can be assigned, about four billion. IPV6 will allow us to assign a network address to every computer, person, insect, microbe, and particle of gravel^[*] on the planet and still have plenty left over. It will also greatly ease the job of routing Internet data around the globe.

  ^[*] I'm exaggerating a bit. There actually aren't enough IPV6 addresses to include all of the gravel.

  Windows XP supports IPV6, although it's not widely used at present.

Network Adapters

Earlier in the chapter, when I described UTP and coaxial cable media, I described the popular physical media and data link protocols used in LANs, and mentioned the Ethernet and Token Ring data link protocols. When you buy a network adapter, they come lumped together: You're buying a piece of hardware that performs both data-link and physical transport functions.

At the physical level, network cards send packets of data through their physical medium from one card to another. Network cards have two ways of sending data: unicast, which sends data directly from one card to another specific card, and multicast or broadcast, which sends the same data packet to every card on the network. Each network card has an address (like a phone number) that is actually built right into the hardware of the network card. It is called the physical network address or media access control (MAC) address.

When a unicast data packet is addressed directly to a MAC address, only the one intended computer receives and examines the data. When a broadcast is made, every computer receives the packet. When a packet arrives by either means, the network adapter uses a hardware interrupt to inform Windows that data has arrived. Windows reads the data out of the network card and passes it up to the next higher layer in the protocol stack to be examined and acted upon.