The Lowdown on Ethernet
Ethernet is the most common of many ways of allowing computers to talk to each
other. The signals are broadcast over the cable, which is hardly surprising as the system was
developed from Packet Radio by Digital, Intel and Xerox; Intel made the chips, and DEC did the
marketing-it was known as DIX Ethernet, after the manufacturers' initials, hence the connector of
the same name. The IEEE 802.3 standard was based on this, where 802 means February 1980 (in case
you ever wondered).
Stations on the network can send messages whenever they want, with no
precedence or order. Although all stations may receive every transmission sent, only the correct
one will respond. It's called Ethernet because it's not meant to be tied to any particular
medium; Ether was supposed to be a hazy word. It's typically used in a bus configuration, where
each computer is attached to a thin coax cable, terminated at each end with a small resistor (50
ohms) that absorbs unwanted signals so they don't get counted twice. This version, called CNET
(or cheapernet), was developed by 3Com Corporation in 1981.
With Ethernet, messages are
split up into packets and sent round the system in short bursts separated by comparatively long
idle periods, each with its address tacked on the front, and being reassembled at the receiving
terminal.
Some errors (the wrong address, for example) mean that the message is totally
ignored. In practice, there is a short delay before any error is detected so there must be a
minimum packet size, otherwise transmission could finish before mistakes are found.
In
other words, minimum packet size and maximum path length mean that even if two machines at each
end begin transmission at the same time, a collision will be sensed before transmission is
finished. A workstation transmitting must listen long enough to assume the packet arrived safely;
a 512-bit frame is specified in the standard, so it will travel 2,500m before the workstation
finishes processing it, after which a collision may arise; in other words, while transmitting,
the workstation knows it has control of the line. Timing and synchronization issues, rather than
signal strength (though this is important) are why cables have to be of specific lengths, even
with Ethernet. The cable works like a telephone party line, in that only one station can transmit
at any time; all stations can transmit freely, but if a collision occurs between packets, the
culprits back off for a short time and then retry.
The most commonly used system for this
is Carrier Sense Multiple Access (CSMA), which comes with either Collision Avoidance (CA) or
Collision Detection (CD) procedures-the former has slightly less overhead, which is why Apple
used it for its own PCs. Being slow, it had enough to contend with.
Looked at in more
detail, the line is sensed for activity by the originating sender and is "acquired" if
nothing is happening. Line voltage drops significantly if two stations transmit at the same time,
and the first station to notice this sends a high voltage jamming signal around the net to
signify a collision. The stations trying to transmit then back off for a random time interval,
which is doubled if it happens again. After 16 attempts an error condition is reported.
This is an article from
Phil Croucher, author
of
Communications and Networks.
Phil has a way of explaining in "plain" English. The information
is well presented and is well above A+ standard.
|