Ethernet networks can be extended by using a device called a repeater. An Ethernet repeater is a physical layer device with two or more Ethernet ports. These ports can be of any type - AUI (Thick), BNC (Thin), RJ-45 (10 Base-T), or fiber optic (10 Base-FL) in any combination.
The basic function of a repeater is to take the data recieved on any port and immediately forward it to all other ports. In the process of forwarding the data, it is also retimed and amplified to eliminate any distortion which may have been introduced after the signal was originally transmitted.
Repeaters also provide a function called partitioning. If a repeater detects many collisions originating on one of its ports, it assumes that a fault has occured somewhere on that segment and isolates it from the rest of the network. This feature is designed to prevent network faults on one segment from propagating through the entire network.
Although repeaters are great pieces of equipment, they do have some drawbacks. The biggest drawback is that they introduce a small amount of delay in the time it takes a signal to fully propagate through a network. All Ethernet networks use an access protocol called CSMA/CD (Carrier Detect Multiple Access With Collision Detection). In a nutshell, CSMA/CD means "Don't speak while someone else is speaking, and if you do, detect it and stop.".
In order for CSMA/CD to work properly, it needs to be able to detect when a collision occurs. The way it does this is by monitoring what data is actually on the network and comparing it with the data it is supposed to be transmitting. If it detects any difference, it assumes a collision occured and immediately stops sending. CSMA/CD then waits a random amount of time and tries again.
There is the flaw in CSMA/CD which limits network size. The bits sent do not arrive instantly at all points in the network, rather it takes a small amount of time for the signal to travel through the wires and any repeaters on the network. This time can be quantified, and is called Propagation Delay. If the Propagation Delay between the sending device and the most distant device on the network from it is greater than half the smallest frame size allowed then CSMA/CD can no longer properly detect collisions and data can be lost or garbled on the netwok.
The designers of Ethernet did a considerable amount of mathematical calculations and determined that the maximum Propagation Delay constaints could be met by a network consisting of no more than four repeaters and five cable segments. They also found that only three of the five allowed cable segments could have devices attached to them. These constaints are commonly referred to as the "5-4-3" rule.
Legal Ethernet Network
We need to make a very important distinction regarding how to count repeaters. We do not count the total number of repeaters on the network, only the number of repeaters between any two points on the network. For example, the network shown in Figure 1 has more than four repeaters, but it does not have any path between any two devices which exceeds the 5-4-3 Rule.
In Figure One we have drawn the two most distant nodes from one another on the network and labled them A and D. The path the signal takes is drawn in green. Each device marked "Hub" is a repeater. Analyzing the drawing shows that between A and D is a total of three (3) cable segments and two (2) repeaters. This network configuration is legal and should work with no problems whatsoever.
Network Exceeds Repeater Count
Figure Two shows a poorly designed network which exceeds the repeater count rules. In this network, we have a total of five repeaters cascaded in a tree topology. We have marked the two most distant nodes from one another as A and B, and have drawn the path the signal must take between them in red. Analyzing the path shows that between A and B is a total of five repeaters and six cable segments. This network will most likely have a very large number of collisions, or possibly not work at all.
In order to troubleshoot or change a network, it is very important to have a detailed drawing of what is in place. Therefore, the first step in determining the repeater count on a particular network is to prepare a detailed drawing showing all repeaters and cable segments.
Once the network diagram is complete, simply find two nodes which seem to have the most cable segments and repeaters between them. Draw a colored line through the network between the nodes, as we have done in Figure One and Figure Two above. The next step is to simply count the number of repeaters and cable segments between the nodes. If the total number of segments beteen them is less than or equal to five (5), and the total number of repeaters between them is less than or equal to four (4), and between the two points not more than three segments have devices other than the repeaters we are going through attached, then and only then is the network legal. If at any point we violate the 5-4-3 rule, then we need to reconfigure the network to fix things.
One point which is very important to keep in mind is that every port on a bridge or Ethernet switch is on a separate Collision Domain or physical network. A Collision Domain is the collection of devices which can cause a collision among each other. Therefore, the repeater count "starts over" at each bridge or switch port. Note that even though all the switch or bridge ports are separate physical networks they will be one logical network so far as the operation of the system is concerned.
If a network can not be grown large enough with repeaters, then use a bridge or switch to break it up into separate small networks. An example is Figure Two above. If we took the top hub (repeater) in that drawing and replaced it with a bridge or switch, each of the new physical networks would be within the 5-4-3 rule and the network would then be legal.