![[EdCert previous]](/usail/edcert/images/prev.gif)
![[EdCert next]](/usail/edcert/images/next.gif)
![[EdCert top]](/usail/edcert/images/top.gif)
Note:to facilitate faster loading of EdCert documents, this page was borrowed from another source.
Document version 2.5
Copyright 1994 by Charles Spurgeon (c.spurgeon@utexas.edu). This document may be freely redistributed in its entirety provided that this copyright notice is not removed. It may not be sold for profit or incorporated in commercial documents without the written permission of the copyright holder.
4.0 Thick Ethernet -- Type 10BASE5
5.0 Thin Ethernet -- Type 10BASE2
6.0 Twisted-pair Ethernet -- Type 10BASE-T
7.0 Fiber Optic Ethernet -- Types FOIRL and 10BASE-F
8.0 Universal 15-pin Connector
9.0 Configuration Rules: Transmission System Model 1(1)
10.0 Non-standard Ethernet Equipment
11.0 Network Design Guidelines
There are several LAN technologies in use today, but Ethernet is by far the most popular technology for departmental networks. The vast majority of computer vendors provide equipment with Ethernet attachments, making it possible to link all manner of computers with an Ethernet LAN. Because of this widespread use there is a large market for Ethernet equipment, which helps keep the technology competitively priced. The ability to link a wide range of computers using a vendor-neutral network technology is essential in a university environment. For these and other reasons the UT Networking Services group recommends Ethernet technology for use on UTnet.
From the time of the first Ethernet standard, the specifications and rights to Ethernet technology have been easily available to anyone who wished to build Ethernet equipment. This openness resulted in a large Ethernet market, and is one reason Ethernet is so widely implemented in the computer industry today. The specifications for Ethernet were first published in 1980 by a multi-vendor consortium that created the DEC-Intel-Xerox (DIX) standard. Ethernet technology was then adopted by the 802 committee of the Institute of Electrical and Electronics Engineers (IEEE).
The IEEE standard was published in 1985, and its formal title is " IEEE 802.3 Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications." This standard provides an "Ethernet like" system based on the original DIX Ethernet technology. All Ethernet equipment since 1985 is built according to the IEEE 802.3 standard, which is pronounced "eight oh two dot three."
Ethernets can be linked together to form extended networks using devices called bridges and routers. Bridges can be used to link multiple Ethernets within a department to support more computers. Routers are used on UTnet to provide a campus-wide backbone network that spans multiple buildings. While individual Ethernets in a campus LAN system may only support dozens of computers, the total system of UTnet Ethernets linked with bridges or routers supports thousands of machines.
Access to the shared medium is determined by the medium access control (MAC) mechanism embedded in each station interface. The media access control mechanism is based on CSMA/CD, and functions somewhat like a dinner party in a dark room. Everyone around the table must listen for a period of quiet before speaking (Carrier Sense). Once a space occurs everyone has an equal chance to say something (Multiple Access). If two people start talking at the same instant they detect that fact, and quit speaking (Collision Detection.) The CSMA/CD mechanism is invoked for every transmission on the network. The mechanism is designed to enforce fair access to the shared medium so that all stations get a chance to use the network.
If two stations happen to transmit at the same instant their signals collide, the stations are notified of the collision, and they reschedule their transmission. To avoid another collision, the stations involved each choose a random time interval to schedule the retransmission of the collided frame.
If repeated collisions occur for a given transmission attempt, then the stations begin backing off by expanding the interval from which the random retransmission time is chosen. Repeated collisions indicate a busy network. The backoff process, formally known as "truncated binary exponential backoff," provides an automatic method for stations to adjust to traffic conditions on the network.
Computers attached to an Ethernet send application data to one another using high-level protocol packets, which are carried in the data field of Ethernet frames. The system of high-level protocols and the Ethernet system are independent entities that cooperate to deliver application data between computers. A given Ethernet system can carry several different kinds of high-level protocol data. The Ethernet is simply a trucking system that carries packages of data between computers; it doesn't care what is inside the packages.
Each computer on the LAN is equipped with an Ethernet interface which is connected to the media system. For the Ethernet media access control system to work properly, all computers must be able to respond to one another's signals within a specified amount of time. To ensure that every computer can hear the network signals within the specified time, the maximum round trip travel time of signals on the shared Ethernet channel must be limited.
The longer a segment, the more time it takes for a signal to propagate over it. To ensure that the round trip propagation timing limits are met, each media variety has maximum segment lengths defined in the standard. The configuration guidelines for Ethernet provide rules for combining these segments so that the correct signal timing is maintained for the entire network system.
If the specifications for media segment lengths and the configuration rules for combining segments are not followed, then computers attached to the Ethernet system may not hear one another's signals within the required time limit, and could end up interfering with one another.
Therefore, the correct operation of an Ethernet depends upon a media system that is built according to the rules for each media type. More complex systems with multiple segment types must be built according to the multi-segment configuration guidelines provided by the IEEE for combining segments. This guide describes the basic rules for each media type, and also contains the multi-segment configuration rules from the standard.
Now that we've taken a quick look at how Ethernet works, let's see what the Ethernet media varieties look like. But first, we need to learn some jargon.
Starting at the right hand side of the figure we find the physical medium, which is used to carry Ethernet signals between computers. This could be any one of several Ethernet media types, including thick or thin coaxial cable, twisted-pair cable and fiber optic cable. Connected to the medium is the medium dependent interface, or MDI. This part of the standard describes the piece of hardware used for making a direct physical and electrical connection to the medium.
In the case of thick Ethernet, the most commonly used MDI is a type of clamp that is installed directly onto the coaxial cable. For twisted-pair Ethernet, the MDI is an RJ-45 telephone-style jack that provides a connection to the four twisted-pair wires used to carry network signals in the twisted-pair media system.
The next device is called the medium attachment unit, or MAU. This device is called a transceiver in the original DIX Ethernet standard, since it both TRANSmits and reCEIVEs signals on the medium. The medium dependent interface is part of the MAU, providing the MAU with a direct physical and electrical connection to the medium.
Following the MAU is the attachment unit interface or AUI. This is called a transceiver cable in the DIX standard. The AUI provides a path for signals and power carried between the Ethernet interface and the MAU. The AUI may be connected to the Ethernet interface in the computer with a 15-pin connector. The computer itself is defined as data terminal equipment (DTE) in the IEEE standard. Each DTE is equipped with an Ethernet interface that performs the medium access control (MAC) functions.
And there we have it: the DTE contains an Ethernet interface which forms up and sends Ethernet frames that carry the data between computers attached to the network. The Ethernet interface is attached to the media system using a set of equipment that includes an AUI and a MAU with its associated MDI.
The MAU and MDI are specifically designed for each media type used in Ethernet. Coaxial MAUs differ from twisted-pair MAUs, for example, both in the technology used for the actual connection to the media (MDI), as well as the method used for sending Ethernet signals over the media and for detecting collisions.
Notice that in the figure above there are two kinds of DTE configurations shown -- one with an external MAU and one with an internal MAU. With an external configuration the DTE contains only an Ethernet interface, and the AUI and MAU are both located outside the DTE. This is how a DTE looks when connected to a thick coaxial system using an external AUI cable and MAU.
However, it's also possible for the MAU and AUI to be part of the network interface inside the DTE, with the only exposed device being the MDI that connects directly to the network media. This is the type of connection made in the thin coax and twisted-pair media systems. In this case, the AUI is nothing more than a set of wires on the interface board that link the Ethernet chips together.
To help make more sense of this alphabet soup let's look next at the Ethernet media types. We will also show a computer connected to segments of each media type. It should be emphasized that this is just a brief survey, and the descriptions of each media type do not contain all the i nformation you need to correctly build large media systems.
The most popular attachment mechanism (MDI) for a 10BASE-5 MAU is sold by AMP Corporation, and consists of a metal and plastic clamp that makes a direct physical and electrical connection to the coaxial cable. This clamp is also called a transceiver tap, since to install the clamp you must drill a hole into the thick coaxial cable in a process known as tapping the cable. Since this clamp may be installed while the network is active, it is also called a "non-intrusive" tap.
Another, much less popular, form of thick Ethernet MDI consists of a tap composed of two type-N coaxial cable connectors. Installing this tap requires cutting the thick coaxial cable, installing N connectors on each cable end, and then installing the tap as a sort of "barrel" connector in-line with the coaxial cable. Cutting the cable halts the operation of the network, earning this approach the label of " intrusive tap." Thick Ethernet MAUs are equipped with a male 15-pin connector to provide an attachment for the AUI cable. This connector has two locking posts, providing an attachment point for a sliding latch connector.
The standard AUI cable is a relatively thick wire (0.4 inch diameter) that may be up to 50 meters (164 feet) long. "Office grade" AUI cables are thinner (approximately 1/4 inch) and more flexible. Office grade AUI cables also have higher signal loss than standard AUI cables and consequently must be limited in length. One vendor of office grade cables rates them as having four times the amount of signal attenuation as standard cables, and only sells them in two and five meter lengths.
The IEEE standard requires that individual segments be connected together with Ethernet repeaters. The repeater is a signal amplification device that keeps the system operating correctly by cleaning up and amplifying the signals that it repeats from one segment to the other. The repeater also has circuits that ensure that collisions that occur on any segment are propagated onto all other segments to which the repeater is attached. By doing this the repeater makes all segments function as though they were a single big segment, or what is known as a single Ethernet " collision domain." This makes it possible for computers attached to any segment in a system of Ethernet segments linked with repeaters to hear the same signals and to operate as a single LAN.
A thick coaxial segment is known as a "mixing segment" in the multi-segment configuration guidelines. A mixing segment is formally defined as one which may have more than two MDI connections.
Thin Ethernet cable sections must be equipped with male BNC connectors at each end. Segments may be a maximum of 185 meters in length, and not 200 meters as the rounded-up "2" in the shorthand identifier might lead you to believe. The standard requires that multiple segments be linked with repeaters.
The thin Ethernet coaxial segment is defined as a mixing segment, since it can support more than two MDI connections. You may have up to 30 MAUs connected to each thin Ethernet segment. Each repeater connection requires a MAU, and must be counted toward the total of 30 MAU connections per segment. Since thin coaxial cable has higher signal attenuation than thick coax, the limit of 185 meters of cable helps ensure that signal losses are held to acceptable limits. The standard also recommends using high quality BNC connectors with low resistance gold plated center conductors.
The limit on the number of connections, and the recommendation of low resistance connectors is intended to help reduce the DC resistance caused by the coaxial connectors used in a thin Ethernet system. This, in turn, helps ensure that the total DC resistance of the segment is kept low enough so that the essential collision detect mechanism continues to work properly.
There are no special MAU spacing rules in the thin Ethernet media system. However the specifications state that the pieces of coaxial cable used to build a thin Ethernet segment may be no shorter than 0.5 meters (1.64 feet) in length. This effectively limits the minimum spacing between MAU connections to 0.5 meters.
Notice that the BNC Tee is connected directly to the BNC MDI on the interface. The standard notes that the length of the "stub" from the BNC MDI on the interface to the coaxial cable should be no longer than four centimeters, to prevent the occurrence of signal reflections which can cause frame errors.
There are twisted-pair Ethernet cable testers available that allow you to check the electrical characteristics of the cable you use, to see if it meets the important electrical specifications. These specifications include signal crosstalk, which is the amount of signal that crosses over between the receive and transmit pairs, and signal attenuation, which is the amount of signal loss encountered on the segment. The transmit and receive data signals on a twisted-pair segment are polarized, with one wire of each signal pair carrying the positive (+) signal, and the other carrying the negative (-) signal. When connecting a twisted-pair interface to a repeater hub this polarity must be preserved, so that the positive terminal on one end of the segment is connected to the positive terminal on the other end.
However, when you are wiring multiple segments in a building it's much easier to wire the cable connectors "straight through" and not worry about whether the wires in the jumper cables or other twisted-pair cables in your building have been correctly crossed over. The way to accomplish this is to do the crossover wiring inside the multiport repeater hub. The twisted-pair Ethernet standard recommends this approach, and states that each port of the hub that is crossed over internally should be marked with an "X." Twisted-pair MAUs send a special link pulse to one another over the twisted-pair segment when the segment is idle. Vendors can provide a link light on the MAU and if the link lights on both MAUs are lit when you connect a segment, then you have an indication that the segment is working correctly.
The more typical installation uses multiport repeaters, also called "hubs" or "concentrators," to provide a repeater connection between a larger number of link segments. You connect the MAU in the Ethernet interface in your computer to one end of the link segment, and the other end of the link segment is connected to the MAU in the repeater hub. That way you can attach as many link segments with their associated computers as you have hub ports, and the computers all communicate via the repeater hub.
In any twisted-pair Ethernet system with more than two computers, you need a multiport repeater hub to connect the individual segments together, and a five-port hub is shown in the figure. Four of the ports are equipped with twisted-pair MAUs and twisted-pair RJ-45 jacks as MDIs. The fifth port may be connected either to a thin Ethernet segment, or to an outboard MAU using the 15-pin AUI connector.
A common error when connecting a computer to a twisted-pair segment is to use the widely available "silver satin" patch cable typically used to connect telephones to the telephone jack on the office wall. The problem is that the silver satin patch cable for telephones does not have twisted wire pairs in it, and the lack of twisted pairs results in excessive signal crosstalk and "phantom collisions." This occurs because collisions are detected in twisted-pair Ethernet by the simultaneous occurrence of signals on the transmit and receive wire pairs, and excessive crosstalk can trigger the collision detect circuit. This problem can be avoided by using only twisted-pair patch cables rated for use in twisted-pair Ethernet systems to make a connection between the MAU in the computer or the hub and the rest of segment.
To deal with this and other aspects of fiber optic Ethernet, a new set of fiber optic media standards, called 10BASE-F, was developed. This new set of standards includes revised specifications for a fiber optic link segment that allow direct attachments to computers. The 10BASE-F specifications include the following three segment types.
You can attach a 15-pin AUI connector to a thin Ethernet segment, for example, by using an external MAU equipped with a thin Ethernet BNC MDI. The MAU with its BNC MDI is attached directly to a BNC Tee on the thin Ethernet coax, and the 15-pin AUI connector on the MAU is connected to the 15-pin AUI connector on the Ethernet interface with an AUI cable. With a small enough MAU you can even eliminate the AUI cable, and connect the 15-pin connector of the MAU directly to the 15-pin connector on the Ethernet interface of the computer.
Now that we've seen what the Ethernet media varieties look like, let's look at the guidelines used for building a multi-segment Ethernet system with these varieties. The next part of this guide describes one of the models provided by the IEEE for multi-segment configuration.
In the rule-based configuration model shown here, a set of multi-segment configuration rules are provided for combining Ethernet segments based on conservative calculations for the components involved. You shouldn't let the fact that these configuration rules are based on conservative calculations lead you to believe that you can bend the rules and always get away with it. There isn't a lot of engineering margin left in maximum-sized Ethernets, despite the allowances made in the standards for manufacturing tolerances and equipment variances. If you want guaranteed performance and reliability, then you need to stick to the published guidelines.
The multi-segment configuration rules are as follows:
a. The maximum allowable length of any inter-repeater fiber segment shall not exceed 1000 m for FOIRL, 10BASE-FB, and 10BASE-FL segments and shall not exceed 700 m for 10BASE-FP segments.
b. The maximum allowable length of any repeater to DTE fiber segment shall not exceed 400 m for 10BASE-FL segments and shall not exceed 300 m for 10BASE-FP segments and 400 m for segments terminated in a 10BASE-FL MAU.
c. There is no restriction on the number of mixing segments in this case. In other words, when using three repeater sets and four segments, all segments may be mixing segments if desired.
While the configuration guidelines emphasize the maximum limits of the system, you should beware of stretching things as far as they can go. Ethernets, like many other systems, work best when they are not being pushed to their limits.
By building and selling standard Ethernet equipment, vendors can be assured that their devices will operate correctly when attached to any properly configured Ethernet system. As long as your Ethernet system is built using devices and media systems that fully comply with the standards you can use the configuration models to verify that your Ethernet system will operate correctly.
However, there are some non-standard devices designed by vendors to be used in the packet transmission path of an Ethernet media system. By being part of the packet transmission path these devices end up being part of the signal timing that is essential for correct Ethernet operation.
Since these devices are not in the standard and are not covered by any configuration guidelines developed by the IEEE, it's difficult to state exactly what the impact of using such a device will be on your network. It may work, or it may not, depending on the total size of your network, the number of computers attached to each network segment, etc. The performance of non-standard devices varies, and each vendor typically has their own special rules for the operation of their non-standard device. The most commonly used non-standard devices are the multiport transceiver and the media converter.
The problem arises because the thick Ethernet standard requires that each MAU attachment be separated by 2.5 meters of cable from the next MAU attachment. This meant that when you needed to connect a number of machines located in the same space to the network, you had to coil up enough thick Ethernet coax in a wiring closet or under a machine room floor to provide cable to meet the 2.5 meter MAU spacing requirement. By providing several (usually eight) 15-pin AUI connectors in a single multiport transceiver, vendors made it easier to connect groups of computers to thick Ethernet.
However, each multiport transceiver adds a certain amount of delay and other effects to the signals that pass through it, and these effects may vary depending upon which vendor built the multiport transceiver. Since the multiport transceiver is not defined in the IEEE 802.3 standard the extra bit times of delay and other effects it may cause are not included in the configuration guidelines, and your system cannot be verified using the IEEE guidelines. If you use multiport transceivers you should read the vendor's configuration guidelines and follow them carefully. Even then you may find that multiport transceivers may not perform well in large networks, or when attached to stations with maximum-length AUI cables.
Media converters are designed to link segments together inexpensively without using a repeater. While they provide some of the signal amplification functions of a repeater, they do not contain the more expensive circuits used by a repeater to retime signals, rebuild the preamble on the Ethernet packet, partition (isolate) the segment in case of errors, and so on.
The lack of these more expensive circuits explains why media converters were a lower-cost approach to linking segments than repeaters. However, the cost differential between media converters and true 802.3 repeaters has been dropping ever since low-cost repeater chips became more widely available due to the popularity of the twisted-pair Ethernet system.
Therefore, there is little economic reason to use a media converter to link segments. This is especially true when you consider that any Ethernet system that includes media converters cannot be evaluated using either configuration model, since the media converter is not part of the standard set of equipment defined in the Ethernet specifications and included in the configuration rules.
To make sure that your network meets the specifications in the standards and to make it possible to evaluate your network using the configuration rules, you must use true IEEE 802.3 repeaters for all segment interconnections. If in doubt when buying a device that links segments together, ask the vendor to verify that what they are selling is a true IEEE 802.3 repeater, and that it meets all of the specifications in section 9 of the 802.3 standard.
The media descriptions shown in this guide can provide a useful overview, but they cannot provide the detailed information required for larger networks. Installing a small Ethernet can be as simple as buying a twisted-pair hub and some patch cables, for example, and connecting all of your computers to the hub. But in larger systems, the issues of structured cabling systems, what kind of media to use, and which media system will best provide for future growth, can be much more complex.
With regard to the configuration guidelines shown in this guide, note that while the guidelines describe how far you can stretch things, this should not be taken to mean that a good network design should push things to their limits. The design of a multi-lane highway can be a useful analogy, since a highway is somewhat like a LAN in that it is a multiple access system with a "shared channel" whose traffic increases and decreases over a 24 hour period.
When highway engineers design a multi-lane limited access highway they calculate the maximum number of vehicles that can be accommodated given the number of lanes and entrance and exit ramps, topology of the roadway, and so on. Like a LAN, a highway system has a theoretical maximum performance, but you do not want to push the system to its limits. After all, a highway loaded with bumper-to-bumper traffic during the homeward-bound commute is still operating within design limits, but no one is very happy with it.
In much the same way a LAN stretched to its limits with a large number of computers on maximum-length segments can end up loaded to its capacity and still be within the specifications and working properly, but no one will be happy with it. When you design a LAN system, you should not focus on what you can get away with or how far you can push the system. Instead, you should consider how many machines you need to support, how much traffic they will generate, and then build a system that will be able to accommodate the load without serious congestion.
Each network design is a special case, since every group has a different mix of computing equipment, and different computing requirements. In general, the UTnet Networking Services group recommends that conservative design practices be used to help deal with network management issues and network traffic growth. Your network designs should emphasize modular cabling systems and network topologies that can be easily reconfigured and upgraded when traffic growth demands more bandwidth.
We typically recommend that campus groups use twisted-pair Ethernet for new installations, based on twisted-pair hubs centralized in one or more wiring closets. Locating equipment in a wiring closet reduces the number of places you must visit when tracking down a problem. Also, as new network equipment becomes available, you can upgrade your centralized hub equipment as required to improve the capabilities of your network system. Installing high quality twisted-pair wiring to each desktop is another way to provide a network system that is reliable and easily managed, and that can be upgraded to higher speeds in the future.
We further recommend that campus groups use only standard IEEE 802.3 media types and equipment when building Ethernets, to make sure that their networks meet the configuration guidelines and to provide interoperability. With respect to standards you should beware of vendor claims. Ethernet has become a commodity market, and there are a number of vendors selling their own inventions which are not described in any IEEE specification. This issue is often further obscured by vendor claims that their proprietary technologies are "compatible" with IEEE equipment. When in doubt, ask the vendor to provide the exact IEEE standards and specifications that apply to the equipment they are selling.