Earlier, the section "Network Protocols and the OSI Model" discussed briefly the various network protocols, including Ethernet, the most popular physical layer LAN technology in use today. There are actually several different variations of Ethernet, each of them having a different combination of physical media (wires), number of allowed hosts per segment, maximum length of an individual segment, and transmission speed. Ethernet variants include
10BASE-FLEthernet over fiber
Although variants of the Ethernet protocol differ widely in their required media and transmission, they all operate in basically the same manner.
At the lowest level, data is sent or received as a very low-power electrical current or as pulses of light. Your system must have a network interface card capable of communicating using the appropriate Ethernet protocol and capable of transmitting and receiving information on the physical media being used for the network (using electrical current or light pulses). In most Ethernet-based networks, the media used is twisted pair copper cabling and the information is sent and received using electrical current.
Because Ethernet is essentially a shared media protocol, there must be some method for controlling traffic. Ethernet uses a method called Carrier Sense Multiple Access/Collision Detect (CSMA/CD) to control traffic flow across the physical media. Under CSMA/CD, a network device wishing to transmit must first check to see whether anyone else is already transmitting information on the physical media (Carrier Sense). If the network device does not detect any other transmissions, it can then begin to send information down the wire.
Even though transmissions are extremely fast, occasionally two devices will attempt to transmit at the same time, and with more hosts and a longer shared media this becomes more common. When two or more devices transmit at the same time, a collision occurs. When the devices detect the collision, they immediately stop transmitting and wait for a random delay (Collision Detection).
Knowing how Ethernet functions makes it easier to understand the performance differences between a switch-based and a hub-based network. In a switch-based network, each system is connected to a single media connection, with the other end being connected to the switch. The switch gives each system the impression that it is connected to its own, separate segment. There are only two devices competing for access to the network media (the system and the switch). Therefore, the system is able to send and receive data at a much higher overall speed and efficiency. Internally, the switch connects each system to a high-speed media, sometimes called a backplane that can handle data much faster than any of the individual system connections. In addition, a switch will send only the appropriate packets down the media to a specific system. The packets the switch puts on an individual system connection are the packets that system needs to see.
By comparison, a hub-based network still connects each system to the hub on a single media connection. However, inside the hub all the connected devices are connected to a single media strand, usually operating at the same speed as the individual system connections. With all the devices connected to the same media, they are sharing the same wire and can see all the packets on the network, whether they need to see them or not. Each packet the hub receives is sent out to every system connected to the hub. This increased traffic volume results in a far higher number of collisions than a switched network, making hub-based networks less efficient and less secure than switch-based networks. Though less efficient, hub-based networks are still very popular because they are generally cheaper to construct and maintain.