Previous Table of Contents Next


Down, Down, Down

Chances are that if your Ethernet segments were behaving last week, and the week before, and the week before that, you don’t have installation problems, right? After all, nothing has changed. Let’s take a look at the way an Ethernet segment might change, and the problems that might ensue.

The “growabililty” of Ethernet has been, in the past, its downfall. Although most folks expanding an Ethernet network nowadays will probably use a switch to add a hub or a couple of hubs, in the past, they would just cascade hubs. Adding hubs or extending an Ethernet network can put it outside design specifications. One of those design specifications is called the “3-4-5 rule.”

The 3-4-5 Rule

The Ethernet 3-4-5 rule states the following:

•  You may have three populated subsegments (applies to coaxial only, so forget this one).
•  You may have four repeaters or concentrators.
•  You may have five subsegments in a series—that is, one connecting to the next, as with Christmas tree lights.

Because the 3 rule applies only to coaxial networks, let’s concentrate on 4 and 5. A repeater is a device that regenerates whatever signal is on the wire in order to overcome a wire length limitation (100 meters with UTP). In a nutshell, when you echo a signal down the line, it takes time; too many echoes results in a critical loss of timing.

The 5 rule means that you shouldn’t cascade more than five hubs in a row. (A stackable hub with a proprietary link between hubs counts as only one hub.)

Even when you’re in spec and nothing has been changed, things still break. For example, network cards can go bad and stop listening while they’re transmitting, causing the party line to be tied up. Smart hubs have antijabber circuitry that usually catches this, but even hubs go bad. When smart hubs can’t make up for a bad network card, or when they, themselves, are broken, it’s time to play divide-and-conquer.

If you suspect a physical network problem (for example, a particular segment is really slow), you’ll need to keep the network up while you’re troubleshooting. (People take it badly when you stop them from working just so you can troubleshoot a problem that they do not consider to be a showstopper.) In this case, you should consider putting a two-port switch in place, which will allow you to give a particular hub or group of hubs a separate party line from the rest (see Figure 9.3).

You can get a miniswitch from DataComm Warehouse (www.datacommwarehouse.com), CableExpress (networksnow.com), or any of the mail-order network houses. (As of this writing, NetGear seems to be a reasonably available manufacturer of two-port switches.)

Installation is a snap. Disconnect the cascade cable from the second hub, leaving it plugged into the first hub. Plug that cable into port 1 of the switch. Add an additional cable to the switch in port 2. Connect port 2 of the switch to the second hub. Voila! You’re done.


Figure 9.3  A small bridge divides the fault domain or collision domain in half, yet users can still operate fairly transparently.

Summary

Ethernet is a really simple physical networking technology. However, like all networks, it has hard-and-fast rules for its design and implementation. Although Ethernet has used many different types of physical cabling in the past, the simplest and most reliable cabling today is UTP.

Ethernet is simple to add to, and that simplicity makes it a target for ad hoc additions that can bring the network outside design specifications. This isn’t a big deal with switched networks, because each switch port gives each Ethernet segment (or work-station) its own party line rather than following the rules that apply to larger, cascaded Ethernet networks.

Workshop

Q&A

Q Should I use a switch port for every PC?

A If you’re very wealthy, go ahead. Most folks, however, like to put a small hub on each switch port and save dedicated switch ports for servers.

Q Are there errors other than collisions?

A Collisions cause other errors, such as CRC errors and runts. But on the whole, all errors are collision related.

Q Don’t people run their own CAT-V cabling?

A Nope. Once something is in the wall, it’s really a pain to replace it, so it makes sense to do it only once.

Q The 3-4-5 rule confuses me. Do I really have to learn it?

A Not if you’re putting a small hub on each switch port with no additional repeaters.

Quiz

1.  During normal error-free transmission, there may only be _____ signal(s) on any given Ethernet segment.
A.  5
B.  10
C.  2
D.  1
2.  What is a collision?
A.  When two network cards try to transmit simultaneously
B.  When two network cards develop bad microchips
C.  When a user accesses the wrong database
D.  When a user transmits incorrect data
3.  True or false? Errors are a normal part of Ethernet operation.
4.  True or false? Thinnet, Thicknet, and AUI are the most desirable types of Ethernet you can buy.

Answers to Quiz Questions

1.  D
2.  A
3.  True
4.  False


Previous Table of Contents Next