Ethernet is the most common LAN technology in use today. Ethernet is a family of implementations, which have evolved into faster and more reliable solutions all based on a common technology.

Ethernet was pioneered by Digital Equipment Corporation, Intel, and Xerox and first published in 1980. The IEEE modified it and gave it the specification 802.3. The way Ethernet works is closely linked to its original connection type: A coaxial cable was used to join all the hosts together. This formed a segment. On a single segment, only one host could use the cable at a time; because the wire was coaxial, with one positive conductor and one negative conductor, it created a single electrical circuit. This single circuit could be energized by only one host at a time, or a conflict would result as two hosts tried to talk at once and nothing got through. Much the same thing happens when you and a friend try to send at the same time using walkie-talkies; all that is heard is noise. This conflict is called a collision.

CSMA/CD (Carrier Sense Multiple Access with Collision Detection) is the method Ethernet uses to deal with collisions. When a host wants to transmit, it first listens to the wire to see if anyone else is transmitting at that moment. If it is clear, it can transmit; if not, it will wait for the host that is transmitting to stop. Sometimes, two hosts decide at the same instant that the wire is clear, and collide with each other. When this happens, the hosts that were involved with the collision send a special jam signal that advises everyone on that segment of the collision. Then all the hosts wait for a random period of time before they check the wire and try transmitting again. This wait time is tiny—a few millionths of a second—and is determined by the backoff algorithm. (The backoff algorithm is the mathematical equation a host runs to come up with the random number.) The theory is that if each host waits a different amount of time, the wire should be clear for all of them when they decide to transmit again.

Any Ethernet segment that uses coaxial cable (10-BASE 2, 10-BASE 5) or a hub with twisted-pair cabling is a collision environment.

Collisions have the effect of clogging up a network because they prevent data from being sent. The more hosts you have sharing a wire, and the more data they have to send, the worse it gets. A group of devices that are affected by one another’s collisions is called a collision domain. As networks grew, it became necessary to break up collision domains so that there were fewer collisions in each one. Devices called bridges and switches did this; these devices are covered in Chapter 6, “Basic Catalyst Switch Operations and Configuration.”

It is possible to eliminate collisions altogether if we can provide separate send and receive circuits; this is more like a telephone (which allows us to speak and hear at the same time) than a walkie-talkie. This requires four conductors—a positive and negative pair for each circuit. The use of twisted-pair cabling (not coax), which has at least four conductors (and more likely eight) allows us to create a full duplex connection, with simultaneous send and receive circuits. Full-duplex connections eliminate collisions because the host can now send and receive simultaneously.

Modern Ethernet is fast, reliable, and collision free if you set it up right. Speeds of up to 10 gigabits per second are possible with the correct cabling.

Table 1.1 summarizes some of the different Ethernet specifications, characteristics, and cable types. This is not all of them, just an idea of how far Ethernet has come.

A leased line refers to a connection that is installed and provisioned for the exclusive use of the customer. Essentially, when you order a leased line, you get your very own piece of wire from your location to the service provider’s network. This is good because no other customer can affect your line, as can be the case with other WAN services. You have a lot of control over this circuit to do things such as Quality of Service and other traffic management. The downside is that a leased line is expensive and gets a lot more expensive if you need to connect offices that are far apart.

A leased line is typically a point-to-point connection from the head office to a branch office, so if you need to connect to multiple locations, you need multiple leased lines. Multiple leased lines get even more expensive. Leased-line circuits typically run the Point-to-Point Protocol (PPP), High-Level Data-Link Control Protocol (HDLC), or possibly Serial Line Internet Protocol (SLIP). (These protocols are covered in detail in Chapter 7.)

A circuit-switched WAN uses the phone company as the service provider, either with analog dial-up or digital ISDN connections. With circuit-switching, if you need to connect to the remote LAN, a call is dialed and a circuit is established; the data is sent across the circuit, and the circuit is taken down when it is no longer needed. Circuit-switched WANs usually use PPP, HDLC, or SLIP, and they tend to be really slow—anywhere from 19.2K for analog dialup to 128K for ISDN using a Basic Rate Interface (BRI). They can also get expensive because most contracts specify a pay-per-usage billing.

Packet-switched WAN services allow you to connect to the provider’s network in much the same way as a PC connects to a hub: When connected, your traffic is affected by other customers’ and theirs by you. This can be an issue sometimes, but it can be managed. The advantage of this shared-bandwidth technology is that with a single physical connection from your router’s serial port (typically), you can establish virtual connections to many other locations around the world. So if you have a lot of branch offices and they are far away from the head office, a packet-switched solution is a good idea. Packet-switched circuits usually use Frame Relay or possibly X.25.

Cell switching is similar to packet switching; the difference is that with packet-switched networks, the size of the units of data being sent (called frames) is variable. Cell-switched units (cells) are of a constant size. This makes dealing with heavy traffic loads easier and more efficient. Cell-switched solutions such as Asynchronous Transfer Mode (ATM) tend to be big, fast, and robust.