All Ethernet networks use the same principal, there can only be one signal on the network at any one time. All the network cards are using the same frequency and if two signals are on the network at the same time there is a collision. This can be equated to having a radio tuned to one frequency, but receiving many stations. As long as no two stations transmit at the same time there are not problems, but as soon as two stations transmit at the same time on the same frequency the signal is corrupted.

This is a generic diagram showing a network. If hostA sends a signal to hpws1 that signal is sent through the whole network. Since the packets have a destination address of hpws1 all the other systems will ignore it with only hpws1 accepting it. If hostA is sending a file to hpws1 at the same time that two other stations are talking to one another the following happens:

• A collision is detected by the network.
• Both networks cards that were transmitting will stop and wait a random length of time.
• One of them will be the first and send a jam signal through the whole network. This tells all the cards to stop transmitting.
• Then this network cards waits another random amount of time and resends the data.

As you can see a network with a lot of systems or a network with few systems that send large files to each other will have its performace degraded. Actually the average 10Base network peaks out at a speed of 3 megabits per second instead of 10. With high network traffic either by having a lot of systems or large file transfers the speed goes even lower.

If we segment the network into collision domains (CD's) then we can increase the effective speed of the network. By splitting the network into segments, ie. A and B we can isolate the traffic. This is done using an intelligent bridge and nothing has to be done with changing network addresses on the systems.

• An intelligent bridge will learn which systems are on segment A and segment B. It does this by listening to the network traffic and is done automatically within 3-5 minutes.
• If a file is sent between hostA and hpws3 then that signal is not forwarded to segment B.
• If a file needs to be sent from hostA to hpws4 then the signal is forwarded from segment A to segment B.
• This means that hostA can be sending a file to hpws3 and hpws4 can be sending a file to ws2 at the same time without causing a collision.

The higher priced hubs can be segmented into collision domains for a 10BaseT network. This will allow us to increase network performance. Below is an example of how a segmentable hub can separate the Sophis USA network to increase per0 Switched Networkcompany thatcase would speed up the network a great deal. This would also help a company that works with very large files that need to be copied between workstations.

There is an even better way of segmenting a 10BaseT network by using a switch. In the diagram above while file transfers within each segment do not cause traffic on the other segment, traffic within each segment will still cause collisions. A switch works by segmenting the network between systems on the fly, giving each connection between systems the full 10 megabits per second bandwidth. It segments machine to machine.

In this example the switched hub will dynamically segment any two systems into a separate CD giving them full 10 megabits per second bandwidth

• Hpws1 to hpws2, hostA to ws1, ws4 to hpws6. These connections will have the full 10MBS bandwidth.
• Then the following happens, hpws1 to hostA, hpws2 to ws4, ws1 to hpws6. The switch resegments the network instantly giving these connections the full 10MBS bandwidth.

A switched hub would be great for a customer who wants to combine their print and weaving systems on one network and have the ability for a designer (print or weaving) to sit at any open system and work. A switch in their case would speed up the network.

For further questions, please contact Pedro Aguiar.