Any communication system must consist of a sender, a channel and a receiver. In addition, when data is to be transmitted over the system it must be coded in some specific manner and there must be some kind of error control built into the system. However, when computer networks are involved another important factor which must be considered is the methods which have to be employed to ensure that information is transmitted to the correct station. In this context networks can be divided up into two broad areas:

• BROADCAST SYSTEMS - Since transmitted messages are available to all stations the receiver must distinguish being those messages addressed to it and those that are not
• DIRECTED SYSTEMS - A transmitted message is only available to the intended receiver

The features of any particular network depend largely on the way in which the components of the network are connected to one another - the topology.

Tree Networks

This hierarchical topology is one of the commonest in use for Wide Area Networks. Generally the node at the top of the tree is in control of the entire network, although occasionally each node has control of those immediately below it.

The diagram above shows that the tree network is relatively simple to control. However, its structure can lead to a number of problems.

Characteristics

• If the top node is assigned all the tasks then quickly a bottleneck will occur and terminal traffic will increase.
• If the top node is out of operation then the entire network is disabled.
• Additional nodes can easily be added at the bottom of the tree.

Star Networks

Again this topology is generally used for Wide Area Networks, however, some LANS have a Star structure. Usually the central node has the peripherals that are to be shared across the network attached to it, and often it acts as a file and printer server to the rest of the stations on the network. The diagram below shows this:

Characteristics

• Easy to set up and relatively simple architecture
• Software requirements are straightforward and thus there are no complex routing problems
• All traffic must pass through the central node and hence if this node is disabled the network is useless
• Shared resources can only be attached to the central node

Ring Networks

In this arrangement of components the stations are connected by point-to-point links and consequently form an unbroken circle. Usually rings are unidirectional and operate by passing packets of data from one station to the next until the correct destination is arrived at - thus rings are broadcast systems.

When a station receives a packet, it checks the address before retransmitting it. If the packet is for that station, it is marked as having being received, before retransmission and hence when the packet reaches the sender once more, the mark indicates that it has been received. The diagram below shows a typical ring topology:

Characteristics

• Each station is connected to the network via a repeater - responsible for receiving, checking an retransmitting packets
• Most rings have a monitor station which has the task of removing corrupted or unwanted packets from the network
• Ring networks can be extremely fast and reliable - they can continue to operate even if one or more stations fail
• Shared resources can be attached to any station
• If the connection between any pair of stations is lost, the entire network goes down

Well known ring network systems include the Cambridge Ring and the IBM Token Ring.

Bus Networks

In this context all the stations are connected to a single cable, as shown below:

Characteristics

• Stations may be plugged anywhere at any time without affecting the integrity of the network
• The cable must be terminated to avoid echoes
• Only one message can be on the bus at any one moment
• A broadcast system commonly used for Local Area Networks
• Cheap and easy to set up
• If the channel goes down the whole network is useless

The best known bus network is Ethernet - Econet is an Ethernet-type network.

Mesh Networks

The Mesh layout is used in Wide Area Networks and allows multiple routes between hosts, as shown below:

A mesh network where every node is connected to every other node is known as a fully connected network, and is shown below:

Characteristics

• Mesh topologies are relatively immune to bottleneck and component failure
• As a result of the variety of paths traffic can be routed around failed or busy components
• Mesh topologies can be expensive in terms of the cabling costs