Ethernet Networks

Ethernet
One of the key concepts behind Ethernet—that of allocating the use of a shared
channel—can be traced to the pioneering efforts of Dr. Norman Abramson
and his colleagues at the University of Hawaii during the early 1970s. Using a
ground-based radio broadcasting system to connect different locations through
the use of a shared channel, Abramson and his colleagues developed the concept
of listening to the channel before transmission, transmitting a frame of
information, listening to the channel output to determine whether a collision
occurred, and, if it did, waiting a random period of time before retransmission.
The resulting University of Hawaii ground-based radio broadcasting system,
called ALOHA, formed the basis for the development of numerous channel
contention systems, including Ethernet. In addition, the subdivision of transmission
into frames of data was the pioneering work in the development of
packet-switching networks. Thus, Norman Abramson and his colleagues can
be considered the forefathers of two of the most important communications
technologies, contention networks and packet-switching networks.
Evolution
The actual development of Ethernet occurred at the Xerox Palo Alto Research
Center (PARC) in Palo Alto, California. A development team headed by
Dr. Robert Metcalfe had to connect over 100 computers on a 1-km cable. The
resulting system, which operated at 2.94 Mbps using the CSMA/CD access
protocol, was referred to as ‘‘Ethernet’’ in a memorandum authored by Metcalfe.
He named it after the luminiferous ether through which electromagnetic
radiation was once thought to propagate.
During its progression from a research-based network into a manufactured
product, Ethernet suffered several identity crises. During the 1970s, it endured
such temporary names as the ‘‘Alto Aloha Network’’ and the ‘‘Xerox Wire.’’
After reverting to the original name, Xerox decided, quite wisely, that the
establishment of Ethernet as an industry standard for local area networks
would be expedited by an alliance with other vendors. A resulting alliance
with Digital Equipment Corporation and Intel Corporation, which was known
as the DIX Consortium, resulted in the development of a 10-Mbps Ethernet network.
It also provided Ethernet with a significant advantage over Datapoint’s
ARCNet and Wang Laboratories’ Wangnet, proprietary local area networks
that were the main competitors to Ethernet during the 1970s.
The alliance between Digital Equipment, Intel, and Xerox resulted in the
publication of a ‘‘Blue Book Standard’’ for Ethernet Version 1. An enhancement
to that standard occurred in 1982 and is referred to as Ethernet Version 2
or Ethernet II in many technical publications. Although the DIX Consortium
submitted its Ethernet specification to the IEEE in 1980, it wasn’t until 1982
that the IEEE 802.3 CSMA/CD standard was promulgated. Because the IEEE
used Ethernet Version 2 as the basis for the 802.3 CSMA/CD standard, and
Ethernet Version 1 has been obsolete for over approximately two decades, we
will refer to Ethernet Version 2 as Ethernet in the remainder of this book.
Network Components
The 10-Mbps Ethernet network standard originally developed by Xerox,
Digital Equipment Corporation, and Intel was based on the use of five hardware
components. Those components include a coaxial cable, a cable tap,
a transceiver, a transceiver cable, and an interface board (also known as
an Ethernet controller). Figure 3.1 illustrates the relationships among Ethernet
components.
Coaxial Cable
One of the problems faced by the designers of Ethernet was the selection of an
appropriatemedium. Although twisted-pair wire is relatively inexpensive and
easy to use, the short distances between twists serve as an antenna for receiving
electromagnetic and radio frequency interference in the form of noise. Thus,
the use of twisted-pair cable restricts the network to relatively short distances.
Coaxial cable, however, has a dielectric shielding the conductor. As long
as the ends of the cable are terminated, coaxial cable can transmit over
greater distances than twisted-pair cable. Because the original development of
Ethernet was oriented toward interconnecting computers located in different
Ethernet hardware components.When thick coaxial cable is used
for the bus, an Ethernet cable connection is made with a transceiver cable and
a transceiver tapped into the cable.
buildings, the use of coaxial cable was well suited for this requirement. Thus,
the initial selection for Ethernet transmission medium was coaxial cable.
There are two types of coaxial cable that can be used to form the main
Ethernet bus. The first type of coaxial cable specified for Ethernet was a relatively
thick 50-ohm cable, which is normally colored yellow and is commonly
referred to as ‘‘thick’’ Ethernet. This cable has a marking every 2.5 meters to
indicate where a tap should occur, if one is required to connect a station to the
main cable at a particular location. These markings represent the minimum
distance one tap must be separated from another on an Ethernet network.
The outer insulation or jacket of the yellow-colored cable is constructed using
PVC. A second popular type of 50-ohm cable has a Teflon jacket and is colored
orange-brown. The Teflon jacket coax is used for plenum-required installations
in air-handling spaces, referred to as plenums, to satisfy fire regulations.
When installing a thick coaxial segment the cable should be rolled from a
common cable spool or cable spools manufactured at the same time, referred
to as a similar cable lot, to minimize irregularities between cables. Under
the Ethernet specifications when the use of cable from different lots cannot
be avoided, cable sections should be used that are either 23.4 m, 70.2 m,
or 117 m in length. Those cable lengths minimize the possibility of excessive
signal reflections occurring due to variances in the minor differences
in cable produced by different vendors or from different cable lots from the
same vendor.
A second type of coaxial cable used with Ethernet is smaller and more
flexible; however, it is capable of providing a transmission distance only onethird
of that obtainable on thick cable. This lighter and more flexible cable is
referred to as ‘‘thin’’ Ethernet and also has an impedance of 50 ohms. When
the IEEE standardized Ethernet, the thick coaxial cable–based network was
assigned the designation 10BASE-5, while the network that uses the thinner
cable was assigned the designator 10BASE-2. Later in this chapter we will
examine IEEE 802.3 networks under which 10BASE-5, 10BASE-2, and other
Ethernet network designators are defined.
Two of the major advantages of thin Ethernet over thick cable are its cost
and its use of BNC connectors. Thin Ethernet is significantly less expensive
than thick Ethernet. Thick Ethernet requires connections via taps, whereas
the use of thin Ethernet permits connections to the bus via industry standard
BNC connectors that form T-junctions.
Transceiver and Transceiver Cable
Transceiver is a shortened form of transmitter-receiver. This device contains
electronics to transmit and receive signals carried by the coaxial cable.
The transceiver contains a tap that, when pushed against the coaxial cable,
penetrates the cable and makes contact with the core of the cable. Ethernet
transceivers are used for broadband transmission on a coaxial cable and
usually include a removable tap assembly. The latter enables vendors to
manufacture transceivers that can operate on thick and thin coaxial cable,
enabling network installers to change only the tap instead of the entire device
and eliminating the necessity to purchase multiple types of transceivers to
accommodate different media requirements. In books and technical literature
the transceiver, its tap, and its housing are often referred to as the medium
attachment unit (MAU).
The transceiver is responsible for carrier detection and collision detection.
When a collision is detected during a transmission, the transceiver places
a special signal, known as a jam, on the cable. This signal, described in
Chapter 4, is of sufficient duration to propagate down the network bus and
inform all of the other transceivers attached to the bus node that a collision
has occurred.
The cable that connects the interface board to the transceiver is known
as the transceiver cable. This cable can be up to 50 meters (165 feet) in
length and contains five individually shielded twisted pairs. Two pairs are
used for data in and data out, and two pairs are used for control signals in
and out. The remaining pair, which is not always used, permits the power
from the computer in which the interface board is inserted to power the
transceiver.
Because collision detection is a critical part of the CSMA/CD access protocol,
the original version of Ethernet was modified to inform the interface board that
the transceiver collision circuitry is operational. This modification resulted in
each transceiver’s sending a signal to the attached interface board after every
transmission, informing the board that the transceiver’s collision circuitry
is operational. This signal is sent by the transceiver over the collision pair
of the transceiver cable and must start within 0.6 microseconds after each
frame is transmitted. The duration of the signal can vary between 0.5 and
1.5 microseconds. Known as the signal quality error and also referred to
as the SQE or heartbeat, this signal is supported by Ethernet Version 2.0,
published as a standard in 1982, and by the IEEE 802.3 standard. Although
the heartbeat (SQE) is between the transceiver and the system to which it is
attached, under the IEEE 802.3 standard transceivers attached to a repeater
must have their heartbeat disabled.
The SQE signal is simply a delayed response by a few bit times to the
transmission of each frame, informing the interface card that everything is
working normally. Because the SQE signal only flows from the transceiver
back to the interface card, it does not delay packet transmission nor does it
flow onto the network. Today most transceivers have a switch or jumper that
enables the SQE signal, commonly labeled SQE Test, to be disabled. Because
repeaters must monitor signals in real time and cannot use the Ethernet time
gap of 9.6 ms between frames (which we will discuss later in this book), this
means that they are not capable of recognizing a heartbeat signal. It should be
noted that a twisted-pair 10BASE-T Ethernet hub is also a repeater. If you fail
to disable the SQE Test signal, the repeater electronics to include hub ports
will misinterpret the signal as a collision. This will result in the transmission
of a jam signal on all hub ports other than the port receiving the SQE Test
signal, significantly degrading network performance.
Interface Board
The interface board, or network interface card (NIC), is inserted into an
expansion slot within a computer and is responsible for transmitting frames
to and receiving frames from the transceiver. This board contains several
special chips, including a controller chip that assembles data into an Ethernet
frame and computes the cyclic redundancy check used for error detection.
Thus, this board is also referred to as an Ethernet controller.
Most Ethernet interface boards contain a DB-15 connector for connecting
the board to the transceiver. Once thin Ethernet cabling became popular,
many manufacturers made their interface boards with both DB-15 and BNC
connectors. The latter was used to permit the interface board to be connected
to a thin Ethernet cable through the use of a T-connector. Figure 3.2 illustrates
the rear panel of a network interface card containing both DB-15 and BNC
connectors. With the development of twisted-pair-based Ethernet, such as
10BASE-T, modern Ethernet interface boards, which are commonly referred
to as network interface cards (NICs), also include an RJ-45 connector to
accommodate a connection to twisted-wire-based networks.
Cabling Restrictions
Under the Ethernet standard developed by Xerox, Digital Equipment Corporation,
and Intel Corporation, a thick coaxial cable is permitted a maximum
length of 500 meters (1640 feet). Multiple cable segments can be joined
together through the use of repeaters; however, the maximum cable distance
between two transceivers is limited to 2.5 km (8200 feet), and no more
than four repeaters can be traversed on any path between transceivers.
Each thick trunk cable segment must be terminated with what is known as
an N-series connector on each end of the cable. The terminator ‘‘terminates’’
Figure 3.2 Ethernet interface board connectors. The first
generation Ethernet interface boards (network interface
cards) contain both DB-15 and BNC connectors to support
the use of either thick or thin coaxial cable. A second
generation of interface cards included an RJ-45 connector
to Accommodate a connection to twisted-wire-based
networks.
the network and blocks electrical interference from flowing onto what would
otherwise be exposed cable. One N-series connector also serves as a ground,
when used with an attached grounding wire that can be connected to the
middle screw of a dual AC electrical power outlet.
Figure 3.3 illustrates a thick Ethernet cable segment after an installer fastened
N-series plugs to each cable end. This is normally accomplished after
the desired length of coaxial cable is routed to form the required network bus.
Next, an N-series terminator connector is fastened onto one N-series plug,
while an N-series terminator with ground wire is fastened onto the N-series
plug at the opposite end of the cable segment.
In addition, as previously mentioned, attachments to the common bus must
be separated by multiples of 2.5 meters. The latter cabling restriction prevents
reflections caused by taps in the main cable from adding up in phase and being
mistaken by one transceiver for another’s transmission. For the total network,
up to 1024 attachments are allowed, including all cable sections connected
through the use of repeaters; however, no more than 100 transceivers can be
on any one cable segment.