Ethernet (CSMA/CD Bus)
|
Ethernet, the LAN technology which has become the predominant method of connecting computers together, was first envisioned in the 1970s by its inventor, Dr. Robert M. Metcalfe. At first it was an experimental system used for networking computers at the Xerox Palo Alto Research Centre (PARC), and was based on a packet radio system called Aloha. In 1973 this became Ethernet as we know it, and the first commercial Ethernet products were relased by Xerox in 1976.
The original specification for Ethernet was proposed in 1980 by a DEC-Intel-Xerox consortium. Known as the DIX Ethernet standard, this formed the basis for the IEEE 802.3 standard for Ethernet-type networks. IEEE 802.3 is the standard for carrier sense multiple access with collision detection (CSMA/CD) based networks, covering the physical layer and MAC sublayer as in the 802.5 (token ring) and other IEEE LAN standards.
Technology Basics
CSMA/CD is the method by which Ethernet and 802.3 networks mediate access to the LAN. It works by stations monitoring the network for a carrier signal (carrier sense), and if no carrier is detected, a station transmits. If more than one station transmits at the same time (multiple access), stations will detect a collision (collision detection) and wait for a random period before trying again.
Ethernet is the generic term used to describe both the DIX and the IEEE methods for CSMA/CD networks, although there are differences between 802.3 and the original 1980 Ethernet standards.
Frame Formats
The table below shows the different frame formats used in Ethernet and IEEE 802.3 networks. Field lengths (in bytes/octets) are given in parentheses.
Ethernet Format | |||||
Preamble (8) | Destination Address (6) | Source Address (6) | Type (2) | Data (variable) | Frame Check Sequence (4) |
IEEE 802.3 Format | ||||||
Preamble (7) | Start of Frame Delimier (1) | Destination Address (2/6) | Source Address (2/6) | Length (2) | Data (variable) | Frame Check Sequence (4) |
Although Ethernet is the most widely installed type of LAN, it is far from perfect. Congestion problems can be aggravated on a basic Ethernet LAN because all messages are broadcast to every device on the network in order to find the destination station. CSMA/CD is an efficient technology for lightly loaded networks. However, at a high loading (above 40-50%) there is a good chance that another collision will occur on each retry, so response times become erratic and the network becomes unusable.
The theoretical speed of an Ethernet network is 10 Mbps. It may be possible to achieve this speed if only one station is on the network continually transmitting data, however in practice average transmission rates will be roughly 3-4 Mbps with peaks at about 7 Mbps. This is due to contention between stations and re-transmission of frames due to collisions.
Various kinds of switches are increasingly being employed in Ethernet networks in order to minimise these collisions, and thus increase the capacity available for transmission. Switches are generally used to isolate network segments from each other and to pass frames between these segments at high speeds. This allows resources that are shared been a specific group of nodes on a network, such as servers and printers, to have their own dedicated segment of the network. Also, frequent users and high speed workstations can get extra performance without overloading the rest of the network.
Physical Media
A wide range of physical media is supported by the Ethernet standards, all operating at the same bit rate of 10 Mbps. They include the following:
Fast Ethernet and IEEE 802.3u (100Base-T)
Fast Ethernet or 100Base-T [LIN96] [NOR96] is simply a faster version (100 Mbps) of IEEE 802.3 / 10Base-T (twisted-pair media). IEEE 802.3u is the recently ratified standard for Fast Ethernet LANs. A more efficient implementation, due to better encoding schemes, enables higher speeds to be reached. However, CSMA/CD is still used as the access control protocol, effectively limiting the available bandwidth.
More serious objections to CSMA/CD as an access control method become apparent at higher bit rates and with larger distances between stations. If two nodes transmitting at the same time are sufficiently far apart they may both complete transmission without noticing a collision at all, and Ethernet doesn't cope well with this situation. Also, because of the faster speeds involved, the maximum cable distances in 100Base-T are lower than those acceptable for a 10 Mbps network.
This material focuses on LAN technologies based on the control token access method, and to some extent, ring topologies. Therefore the Ethernet information presented here is only an introduction to the topic. Fortunately, there is lots of information about Ethernet LANs in most textbooks on data/computer communications and on the World Wide Web. Some links are given below.
LANs Home | Introduction | Token Ring | FDDI / CDDI | Future LANs | References |
Adrian Catchpole - A.G.Catchpole@comp.brad.ac.uk