How Ethernet Works

Nathan Edwards

We tend to take things for granted when they work exceptionally well. Take Ethernet, for instance; it’s almost magical: Plug a simple cable into a computer, and it can exchange data with another rig—or many others. Peek behind the curtain and you’ll discover a brilliantly simple yet continually evolving networking system.

But let’s clear up one thing first: Ethernet technology doesn’t actually contain ether (including its form as a chemical analgesic, so don’t bother sniffing the cables). Robert Metcalfe, one of the technology’s inventors, copped the name from an early scientific theory (discredited in 1887) that luminiferous ether was a passive substance that permitted the propagation of light. Metcalfe thought the name suitable because the cable used to build a network is a passive medium that permits the propagation of data.

Ethernet’s Origins

The original Ethernet was created at Xerox’s Palo Alto Research Center in the mid 1970s. It allowed all the computers on the network to be connected to a single cable, using a protocol known as CSMA/CD (Carrier Sense Multiple Access with Collision Detect). When one computer wanted to transmit, it would first check to see if any other machines were using the line. If the line was free, the transmitting computer tagged the data it needed to send with a MAC (Media Access Control) address and loaded it onto the network. The MAC address identified the intended recipient so that the machine possessing that unique MAC address would accept the data and all the other machines on the network would ignore it.

If the transmission line was busy, the computer would wait until it detected a lull, but if two machines tried to transmit at the same time, each would react to the collision by waiting a random number of milliseconds before attempting to retransmit. The process was simple, but it was also very limited. Multiple collisions could quickly throttle the performance of a large network, for example, and it was relatively easy to eavesdrop on the network’s traffic—all it took was a fake MAC address. The network wasn’t very robust, either: Damage to any cable in the network could cause the entire system to crumble.

Switched Ethernet

The introduction of the hub enabled larger Ethernet networks. A hub rebroadcasts network traffic to extend the network’s reach, and it eliminates the problem of one damaged cable bringing down the entire network. But the development of the switch was a far more significant improvement to Ethernet topology. The switch inspects the source and destination addresses of every message carried on the network, and it uses this information to construct a look-up table so that it knows which machine is connected to each of its ports.

When the switch receives a packet of data tagged with a specific MAC address, it can transmit the data directly to the port that the recipient machine is connected to. This leaves all the other ports free, and it minimizes the possibility of collisions.

Pack Your Bags

Data travels over an Ethernet network in packets called frames. A frame consists of a MAC header (consisting of the source and destination MAC addresses and the Ether type), the data (or payload), and a CRC checksum. The Ether type identifies which protocol is being transported inside the frame (e.g., Internet Protocol); the CRC checksum detects any alteration that might have occurred during transmission. A jumbo frame is an Ethernet frame carrying more than 1,500 bytes of payload.

The original Ethernet cable had a shielded, coaxial design with a BNC (Bayonet Nut Connection) at each end. This single cable was shared by every computer on the network. Ethernet networks using unshielded twisted-pair wiring were developed in the mid 1980s. This enabled the network to operate in full duplex mode, meaning data could flow in two directions simultaneously. Although the wiring is unshielded, the twist design blocks most interference (provided the cable is not run in close proximity and parallel to electrical wiring). Twisted-pair wiring, which is terminated with 8P8C jacks, is also much less expensive to deploy than coax.

The most common Ethernet cable is known as Category 5 (Cat5). It consists of four pairs of twisted wires inside a single PVC jacket and can support data rates up to 100Mb/s. Cat5 has since been superseded by Cat5e (the “e” stands for “enhanced”), an improved specification that is capable of supporting data rates up to 1Gb/s. Both types of cables can operate at frequencies up to 100MHz and are limited to runs of 100 meters (328 feet), including the length of patch cables used at each end. Ethernet networks based on this technology (combined with a hub or switch) are known as 10BASE-T (speeds of 10Mb/s), 100BASE-T (speeds of 100Mb/s), or 1000BASE-T (speeds of 1Gb/s).

Category 6 (Cat6) cable also consists of four twisted pairs of wires and is backward compatible with Cat5/Cat5e installations, but it features more stringent specifications for crosstalk (interference caused by the signal transmitted on one channel bleeding into an adjacent channel) and system noise. Cat6 cable can operate at frequencies up to 200MHz and is capable of supporting data rates up to 10Gb/s, but it is still limited to runs of 100 meters.

The Wi-Fi standard operates on the exact same principles as wired Ethernet, with the obvious exception that the data is transmitted over the airwaves instead of cables.

Around the web