| |
|
![[Advertisement]](http://66.35.216.90/materials/793/salogo_ad.gif)
|
As processor speeds and performance demands have increased, networking
technology has kept pace. One popular improvement to local networking
technology has been Fast Ethernet, which sports a tenfold increase
in bandwidth. Many people assume that, other than the speed, Fast
Ethernet behaves essentially the same as standard Ethernet. Although
this is mostly correct, there are some important differences in
the way each network can be interconnected.
Ethernet Basics The entire concept of an Ethernet network was standardized by the Institute of Electrical and Electronic Engineers (IEEE) subcommittee 802.3. Ethernet is a multiaccess medium that allows all hosts to communicate equally. It behaves like a bus--any data sent on an Ethernet wire is seen by all devices connected to it. Ethernet addresses are used by the sender to indicate the message's intended recipient. If two hosts on the wire try to send simultaneously, both fail in what is called a collision. Each host then backs off for a random period of time and tries again. The concepts are summarized as Carrier Sense Multiple Access/Collision Detection (CSMA/CD). The original coaxial-cable Ethernet is physically arranged as a bus, with each station tapping directly into the wire. Twisted-pair Ethernets require an active hub, and the physical layout of the interconnections consists of a collection of point-to-point circuits. However, the twisted-pair hub still uses CSMA/CD across its connections simply by repeating signals between them.
Let's Speed Things Up The primary difference between standard and Fast Ethernet is that everything has been clocked up to run ten times as fast. To accommodate this increase, the physical wiring must meet higher standards. The format of the data on the wire (the Ethernet frames) is the same, as is the use of CSMA/CD. The term "bit time" refers to the amount of time needed to transmit one bit, and the bit time for Fast Ethernet is one-tenth that of a 10-megabit Ethernet. But this simple difference has wide-ranging consequences. There are three types of media defined for use by Fast Ethernet. 100Base-TX consists of two pairs of twisted-pair wire, 100Base-FX designates two fiber-optic strands, and 100Base-T4 requires all four pairs of a twisted-pair circuit. Note that no coaxial media is defined for Fast Ethernet. One significant difference between 100Base-TX and its cousin 10Base-T is the wire itself. 10Base-T requires (at a minimum) what is called a Category 3 twisted pair, whereas 100Base-TX requires Category 5 as the minimum twisted-pair cabling. The primary difference between these categories is the number of twists per inch placed in the wiring. Category 5 cables should not be pulled through conduits as Category 3 cables can because of the possibility of stretching the wire and the result of that on Category 5's increased sensitivity to the number of twists per inch. The 100Base-T4 media can use Category 3 wiring by utilizing four pair. It was provided as a way to get faster speeds for installations stuck with lower-grade wiring. Many installations have simply opted for the expense of upgrading the wiring, making T4 installations somewhat rare.
Circuit Lengths Twisted-pair Ethernet circuits are limited by 802.3 to be no more than 100 meters. The circuit is the link between a host and its hub. In 10Base-T, this limitation is imposed primarily because of signal strength. Beyond 100 meters, the signals used on the wire might degrade to an unreliable level. An installation that uses Category 5 wiring for its 10Base-T circuits might not see a degradation in the signal strength as quickly. Many network designers have gotten used to the fact that these circuits can be pushed well beyond their 100-meter limit with no ill effect. By contrast, the constraining factor for 100Base-TX is the roundtrip time of the signal, not its strength. Higher-grade circuits will not shorten the roundtrip time. As a consequence, these circuits cannot be pushed beyond 100 meters without serious operational problems. Wiring standards recommend that the distance from the communications closet to the wall jack be limited to 90 meters. This leaves 10 meters of slack for patch cords and wiring in the box.
Repeaters And Hubs Fast Ethernet uses repeaters to interlink segments, just like conventional
Ethernet. These devices operate at the electrical level (ISO layer 1) and simply
propagate the signal from one port to all others. A hub is nothing more than
a repeater with many ports. Fast Ethernet introduces one wrinkle with T4. There
are two classes of repeaters, named class I and class II. The class designation
of a repeater (including hubs) usually is indicated somewhere on the box with
a symbol like this:
Collision Domains The biggest concern for Fast Ethernet is the size of the collision domain, which is the segment of the network where a collision occurs if any two hosts attached to the segment transmit at exactly the same time. Collision domains are bounded by routers, bridges, and switches. A hub will propagate collisions, so any segments connected to a hub will all be part of the same collision domain. A collision domain must be small enough so that any collision occurring within it has time to propagate to all parts of the domain. The propagation delay is measured in bit times, and no two hosts in a collision domain can be farther apart than 512 bit times. In conventional Ethernet, the collision domain's size can be kept under its limit by following the 5-4-3 rule of thumb: no more than five segments interconnected with four repeaters, and no more than three segments populated with hosts. Don't believe the misconception that the same rules apply to Fast Ethernet. A signal will move through copper or glass at the same rate regardless of its frequency. But because Fast Ethernet is expecting everything to come in ten times as fast, the length of a bit time must correspondingly be one tenth of what it is for conventional Ethernet. Consequently, any collision occurring from a transmission must be much closer to the originating host if it is to be detected in time. A whole new set of rules applies when laying out 100-megabit Ethernets.
Rules Of Thumb Rather than get into all the grubby details of propagation delays across differing media types, there are some conservative rules that will always yield well-behaved networks. These are the rules of thumb that have permeated the industry. The most popular is the 5-4-3 rule mentioned earlier. As with most rules, this one can be stretched or even broken in some situations. But the rule of propagation delay cannot. With 10-megabit Ethernet, no two stations can be more than 575 bit times away. Anything greater runs the risk of creating undetected collisions. For Fast Ethernet, the limitation is 512 bit times. Fast Ethernet does not have a simple rule of thumb like 5-4-3, but it does have a table of worst-case cable-length limitations. Each twisted-pair segment, of course, is limited to 100 meters. Each 100Base-FX segment is limited to 412 meters. Figure 1 displays the maximum recommended distance between any two hosts for different mixes of media types. Now consider the diagram in Figure 2. The repeaters are connected to any number of TX circuits. In the worst case, these segments can be as long as 100 meters each. If the link L between A and B is copper, the total distance between any two hosts in this network is limited to 205 meters. This leaves only five meters for link L. If the link is fiber, it could be 16 meters. Either way, that's not much distance. If the maximum circuit length in the installation is less than 100 meters, that extra distance can be used up between the routers. (But who knows when some future circuit might be installed that is maximal length?) These constraints are tight partly because the rule of thumb is conservative. It is possible to calculate a more accurate propagation delay for the collision domain by finding the longest path through the network and adding the delay incurred by each component along that path, including the connections at each end, every repeater, and a delay per meter of media. The details of this calculation are too involved to publish here, but the Web site referenced at the end of this article has more information.
Switches Limit Collision Domains These distance limitations seem depressing until you remember that a switch completely changes the picture. An Ethernet switch behaves basically like a multi-way bridge, and it defines the edge of a collision domain. The link between a switch and any host is a separate collision domain, and is kept easily within the domain's limits by the 100-meter constraint on circuit lengths. A hub connected to a switch will see the switch as just another host, and the domain will not include other hosts on the switch. This explains the immense popularity of Fast Ethernet switches. Without them, a Fast Ethernet network is severely limited in physical distance.
Equipment Considerations Take care when investigating new Fast Ethernet equipment. First determine if the product is a hub or a switch. If it is a hub, ensure it is a class II hub (unless you have a need for interoperability with T4). Bear in mind that the properties of a repeater prevent it from operating at more than one speed at a time. A repeater or hub operates either at 10Mbps or 100Mbps, but not both. A switch can handle connections of differing speeds, because it naturally has the technology to buffer packets. Recent developments in the marketplace have included a so-called multispeed hub. This device is actually two hubs connected by a bridge. One hub operates at 10Mbps and the other at 100Mbps. Additional hardware determines the speed of a given port and ensures it is electrically connected to the correct hub. Such a box operates in two different collision domains. If you are planning to migrate to Fast Ethernet, remember the following:
If you are
unsure of your cable infrastructure, consider purchasing or renting
a cable tester to make sure the existing circuits meet Category
5 specifications. The tester should include a near-end-crosstalk
(NEXT) test and meet TIA/EIA TSB-67 level II specifications for
testers.
References Spurgeon, Charles, Practical Networking with Ethernet, Boston, MA: International Thompson Computer Press, 1997. Spurgeon,
Charles, Ethernet Web Site, http://wwwhost.ots.utexas.edu/ethernet/.
|
||||||||||||
|
Home | Top | Editor's Choice
|
|||||||||||||
. The class I repeaters incur a larger delay in propagating the signal, but
they can intermix T4 with FX and TX. The class II repeaters have a shorter delay,
but can't intermix the media types. A class II repeater can repeat between TX
and FX, but not T4. Likewise, a T4 class II repeater cannot accommodate TX or
FX.![[Advertisement]](http://66.35.216.90/materials/789/sa-solaris.gif)