A Network Technology Primer

You can break down all the various network technologies available in today's marketplace according to five categories. Therefore, by answering all five of the following questions, you can distinguish any one type of network technology from another:

• What access method, protocol, and topology does the network use?

• How does the network work?

• What are the network's technical pros and cons?

• What types of network media does the network support?

• How business-friendly (cost, availability, and so on) is this network?

In the following sections, we examine the pros and cons of various network technologies just as you must do when you have to apply a certain technology to a particular topology with the goal of best serving the needs of your users.

And the technology contestants are

When you talk about network technologies, you're talking about a specific type of hardware and associated driver software that, when added to a PC, can produce a working network connection. The working part, however, is contingent on the right infrastructure cables, connections, and ancillary equipment (such as bridges or routers) also being available.

For the purposes of this book, there are two primary network technologies (but that's a tremendous oversimplification, as you'll find out later):

• Ethernet

• Token ring

• Other

Okay, you caught us sneaking in a third, catch-all entry (we almost called it miscellaneous ) to give us the opportunity to say something about several of the multitude of other available (but less common) network technologies that we choose not to cover in detail in this book. In fact, recent industry analyses indicate that it's about 75 percent likely that your network uses (or will use) one or both of the first two network technologies mentioned in the preceding list. Thus, even though we cover only a small number of technologies in depth, we cover most networks to some degree.

Meet Ethernet, the most popular network technology

Ethernet is the best known, most widely used, versatile, and readily available network technology around. As such things go, Ethernet has been around longer than most, since the mid-to-late 1970s. Ethernet was the brainchild of Xerox's Palo Alto Research Center (PARC) and later adopted by Digital, Intel, and Xerox (which is why older 15-pin connectors for ThickWire Ethernet are sometimes called DIX connectors). Ethernet has long been a networking commodity, which means that plenty of vendors play in this market and that lots of options and choices are available for this technology.

Ethernet uses the CSMA/CD access method. The sidebar titled "Ethernet: Network bumper cars ," shown later in this chapter, explains what this stuff means in everyday English, insofar as the subject allows which isn't so very far, alas.

KEY CONCEPT 

The easiest way to describe CSMA/CD is like this: "Listen before sending. Listen while sending. If garbage happens, quit sending and try again later."

Ethernet's strengths and weaknesses

Ethernet's strengths are as follows : It's robust and reliable, and it comes in a broad, affordable range of flavors. Ethernet's weaknesses include the inevitability of collisions and the more difficult troubleshooting techniques that a bus network requires. Ethernet's base speed of 10 Mbps (short for megabits per second ) is on the slow end for modern networks, but plenty of higher-speed Ethernet versions are now available. (We give you the goods on these in Chapter 7.)

Ethernet does not perform well for high-traffic applications or when real-time delivery is needed (for video and multimedia), nor does it degrade gracefully when high traffic volumes occur. In fact, with Ethernet's CSMA/CD access method, the effective ceiling on its bandwidth is between 56 and 60 percent of total bandwidth (or between 5.6 and 6.0 Mbps on a 10-Mbps Ethernet). That's the level of use beyond which the increasing probability of collisions often results in network slowdowns or failures.

Ethernet: Network bumper cars

The acronym that describes Ethernet's media access method is CSMA/CD, for Carrier Sense Multiple Access with Collision Detection . (The auditory equivalent of a collision is an echo.) When a collision occurs, you must repeat the transmission. The following list provides a definition for each term in this access method acronym:

• Carrier Sense: Everyone attached to the network is always listening to the wire, and no one can send while someone else is sending. When a message moves across the wire, an electrical signal called a carrier is used. By listening to the wire, a device knows when it's busy, because it senses the presence of the carrier.

• Multiple Access: Any device attached to the network can send a message whenever it wants, as long as no carrier is sensed at the time. This means that multiple senders can (and sometimes do) begin sending at roughly the same time when they think things are quiet and that's why it's called multiple access .

• Collision Detection: If two or more senders begin transmitting at roughly the same time, sooner or later their messages run into each other, causing a collision . Collisions are easy to recognize because they produce a garbage signal that is completely unlike a valid transmission. Ethernet hardware includes collision-detection circuitry that immediately halts transmission when a collision is observed . When a collision occurs, each sender waits a random time interval before listening to the wire to retry its transmission.

 

Warning 

When planning bandwidth consumption for an Ethernet network, use 55 percent of the total bandwidth (5.5 Mbps on a 10-Mbps network, 55 Mbps on a 100-Mbps network, and so on) as the ceiling for usable bandwidth on any network segment. If you plan to consume Ethernet's entire bandwidth when designing a network, you'll be designing a network that's headed for trouble!

There's no shortage of bandwidth available to Ethernet customers today. Most newer Ethernet network interface cards (NICs) are 10/100 designs, which means they can sense whether they're used on a 10-Mbps or 100-Mbps Ethernet network and set their speeds accordingly . Today, Gigabit Ethernet, with an amazing 1,000 Mbps of total theoretical bandwidth, lifts the ceiling on network capacity to new heights but retains compatibility with other Ethernet versions.

All the flavors of Ethernet

Ethernet comes in all the basic flavors. That is, Ethernet runs on any of the major media types twisted-pair, coaxial cable (multiple versions, in fact), and fiber optic and works with both bus and star topologies. One unusual variant 100BaseVG-AnyLAN uses a different access method called demand priority that gives this implementation interesting capabilities. (100BaseVG-AnyLAN is covered in detail in Chapter 7.)

Also, you can easily find Ethernet devices that allow you to mix and match media; so you can use Ethernet to build networks of just about any size and for even the most hostile environments (such as factory floors or engine rooms, where lots of heavy-duty equipment can create major interference).

In addition, Ethernet technologies support some innovative uses of bandwidth, so you'll occasionally hear about varieties such as switched Ethernet and full-duplex Ethernet. The former variety depends on a special kind of device (called a switch, naturally) that allows any two nodes to establish a private end-to-end connection. Therefore, switched Ethernet allows pairs of machines to use the entire bandwidth of the network medium. (This is a great way to squeeze extra life out of 10-Mbps Ethernet systems.) Full-duplex Ethernet is limited to 100BaseVG-AnyLAN and uses two pairs of wires so that machines can send and receive data at the same time, thereby doubling overall bandwidth.

The business end of Ethernet

Despite its age, Ethernet remains the most widespread and popular network technology. Of the major media types available, twisted-pair leads the pack for new Ethernet installations, but a lot of coaxial cable is still in use. Fiberbased Ethernet is usually limited to networks in campus environments, where long distances and electrical interference issues are greatest. However, it's also used in hostile environments or for high-bandwidth applications, including both 100-Mbps and gigabit implementations .

The primary reasons for Ethernet's unshaken popularity are as follows:

• Affordability: Cabling is cheap and interfaces range from $20 for bottom-end NICs to less than $200 for powerful server NICs. Ethernet is not the cheapest of all the network technologies, but it's darn close!

• Freedom of choice: Ethernet supports all types of media, numerous bandwidths, and lots of gear to build hybrid networks. Vendors galore offer Ethernet hardware. For specialized network hardware or media needs, chances are good that some Ethernet variety meets them. If some option isn't available, it's probably on someone's drawing board.

• Experience: Ethernet's longevity means that Ethernet-savvy individuals are easy to find. Also, lots of technical and training material on Ethernet makes expertise relatively easy to build.

• Continuing innovation: At 10 Mbps, basic Ethernet is no speed demon. However, vendors make high-speed network switches for Ethernet that can deliver the entire 10 Mbps to individual connections, and higher-speed Ethernet varieties are readily available and widely used. As bandwidth needs grow, engineers have found ways to increase Ethernet's capabilities to match those needs, as the formalization of Gigabit Ethernet as an IEEE standard (802.3z, in fact) attests.

Tip 

If you're asked to build a new network and no compelling reasons exist to choose another network technology, choose Ethernet because of all the reasons previously mentioned!

Taking on token ring

Token ring has gained a substantial foothold in the marketplace, although it hasn't been around in commercial form as long as Ethernet. Token ring is based on technology refined and originally marketed by IBM, so it's most commonly found in environments where IBM is entrenched. When PCs started taking desktop space away from dumb terminals hooked to IBM mainframes, IBM took action. They developed token ring to tie all those new PCs into their mainframe computers.

Token ring uses a token-passing access method in a collection of individual point-to-point links between pairs of devices arranged in a circular pattern. Point-to-point means that one device is hooked directly to another. For token ring, point-to-point describes the connection between a computer and a hub, which may, in turn , be attached to other hubs or computers. Although the devices used with token-ring networks act like hubs, they are more properly known as multistation access units (MAUs or MSAUs) or controlled attachment units (CAUs). The reason they're not really hubs (and that they're more expensive than most hubs) derives from these devices' capabilities to reconfigure themselves on the fly as nodes enter and leave the network. This is more difficult than simply sensing whether a connection is working and requires more expensive hardware to handle the job.

Token ring is mathematically fair to everyone who participates, and it guarantees that the network isn't overwhelmed by traffic. Token ring is said to be fair because it constantly passes the right to transmit around the network. This is accomplished in the form of a special message called a token . To send a message, a computer must wait until it obtains possession of the token. The token is not released until the message has been delivered (or until it's obvious that it can't be delivered). Everyone gets the same opportunity at using the token.

KEY CONCEPT 

The easiest way to think about how token ring works is as follows: To send a message, your system waits for the token. When the token comes by, if it's not already carrying a message, your system tacks your message to the token, and then sends the token (and the message) on to its intended recipient. The recipient, upon getting a token addressed to it, copies the attached message and passes the token on around the ring. When the token comes back to your system, it strips off your message and sends the token (now empty of cargo) to the next computer downstream on the ring.

Token ring's strengths and weaknesses

Token ring's strengths include equal access for all devices and guaranteed delivery. Token ring works reliably and predictably, even when loaded to capacity. Token ring is available in two speeds: 4 Mbps and 16 Mbps. The older, slower version runs at 4 Mbps. This is 40 percent of the theoretical bandwidth for 10-Mbps Ethernet, but only slightly slower than basic Ethernet's effective speed.

A newer, higher-speed version of token ring runs at 16 Mbps, or 160 percent of basic Ethernet's theoretical bandwidth. It can handle three to four times as much data because it allows simultaneous use of multiple tokens while using 100 percent of total bandwidth. Waiting in the wings is a full-duplex implementation of token ring that works much like switched Ethernet. For higher speeds, a 100-Mbps version of token ring is under construction.

We hear you thinking "If token ring is so great, why buy Ethernet?" Token ring has weaknesses that have less to do with technical considerations and more to do with inflexibility and expense. Token ring's major downside is that it requires the expensive MAUs we discussed in the preceding section. Also, token ring requires that two strands of cable be run from each computer to each hub port. (One for the outbound trip, the other for the return trip.) These requirements add to token ring's expense and reduce the maximum legal distance between computers and hubs. Token ring is also more complicated and requires fancier connectors than Ethernet.

Token ring's many flavors

Token-ring implementations for twisted-pair and fiber-optic media are available, but twisted-pair is by far the most common implementation and is the most likely medium when tying desktops to hubs. Fiber-optic is the cable of choice for spanning longer distances and daisy-chaining MAUs. Only limited amounts of shielded twisted-pair (STP) cabling are used on token-ring networks. Because of individual cable length limitations and maximum ringsize limitations, cabling a token-ring network takes more planning and number crunching than cabling an Ethernet network.

The business end of token ring

From a cost-benefit perspective, there's not enough benefit in token ring's reliability, fairness, and guaranteed performance to offset its higher costs. Today, token ring costs from 75 to 150 percent more than Ethernet without necessarily providing significant performance or reliability advantages.

Although there are two schools of thought on this issue "Forget token ring. Ethernet rules. 'and' We're token-passing fools. What's Ethernet?" we're not about to climb out on either limb. If someone offers you token ring at a price that's too good to pass up or if that's what circumstances dictate you must use, go ahead and use it. Token ring works just fine. However, we don't recommend it as a technology of choice for starter networks because of its expense and complexity.

What other network technologies are there?

If you're thinking, "What else is there?" you may be wondering why this book omits your network technology of choice. We hate to be the bearer of sad tidings, but if you're not using either Ethernet or token ring, you may be networking too hard (or at least networking the hard way). Sorry!

In fact, hundreds of other types of network technologies are in use today. At least one such technology exists for every letter in the alphabet from A for ARCnet to X for xDSL. If you think that such acronym overkill may end up making your head spin, you're not alone. The proliferation of exotic network technologies can be a problem for Windows Server 2003, too. A word to the wise: If you're using an exotic network technology, make sure that Windows Server 203 works with it before you spend any money on the software.

The good news is that Windows 2003 runs over a reasonable subset of network technologies. The bad news is that you'll have to do some research to find out whether what you're using is one of the technologies that Windows 2003 supports. Worse yet, you'll have to work harder to do basic stuff that less exotic network technologies take for granted, and you may have to pass on some sophisticated capabilities as well, such as network-attached printers and other peripheral devices.

Doing without Ethernet or token ring does not necessarily bring with it such harrowing consequences, however. In the following two sections, we do your homework for you and cover a couple of potentially useful network technologies that occur on a reasonably large subset of enterprise networks and are supported by Windows 2003.

Find your way to FDDI

One workhorse network technology found on many networks, especially in campus environments, is called the Fiber Distributed Data Interface (FDDI). FDDI uses a token-passing access method. FDDI cable uses a real ring topology, but consists of two rings. One ring transmits messages clockwise; the other transmits messages counterclockwise. If either ring fails, the other automatically takes over as a backup. Better still, if both rings get cut in the same place watch out for guys with backhoes on your campus the two rings automatically splice together to form a ring that's twice the length of the original ring and still able to function.

FDDI's biggest advantage is reach, now that its 100-Mbps speed is no longer such a big deal. FDDI supports rings as big as 100 kilometers in circumference (that's about 62 miles for non-metric types). FDDI can support as many as 500 active devices on a single ring, which is more than the other network technologies we discuss in detail in this chapter.

On the downside, FDDI requires fiber-optic cables for runs of any length. There's a CDDI (where C stands for copper ), but it doesn't support cable runs longer than 75 feet, so it's impractical except for workstation connections.

Cost is another negative for FDDI. Fiber-optic cables are more expensive to buy and install than other types of cable, and FDDI NICs cost between $700 and $1800 each. FDDI looks good primarily for the central line of a campus network (the backbone, to use the technical term), if Gigabit Ethernet is not an option (or a requirement).

Accelerating to ATM

A rising star in the area of high-speed networking is ATM, an acronym for asynchronous transfer mode. In the last few years , several companies have introduced ATM-based local area network (LAN) equipment. Long-distance telephone companies already use ATM versions that run at 155 and 622 Mbps, and current ATM specifications support speeds of 1.2 and 2.4 gigabits per second (Gbps) as well.

ATM is a fast-switching technology that requires a hub-like switching device and network interfaces for each computer. You can expect to spend at least $2000 per workstation (including allocated switch and fiber-optic cable costs) to bring ATM to your computers. That probably explains why ATM is far more popular as a backbone technology than for networking computers.

Figure 4-3 gives you the skinny on the comparative speeds of the different network technologies (which will help you a little), but the best way to find out what's what is to go online and seek out the collective wisdom that's so readily available. You can check in on any of Microsoft's Usenet Windows 2003-focused newsgroups through news://msnews.microsoft.com; use Microsoft's own online service, Microsoft Network (MSN), at http://www.msn.com; or drop in on any of the numerous Windows 2003-focused mailing lists and ask, "Does my (fill in the name of your network technology here) work with Windows Server 2003?" If you don't have access to online information resources, ask around. With so many users worldwide, you shouldn't have to look far to find someone who can help you.

Figure 4-3: Ranking the speed of networking technologies.