There are many different methods (topologies) for interconnecting two or more digital computers together in a network. This document outlines the advantages and disadvantages of the most common network topologies.
Point to Point
One possible solution would be to provide a transmission route from each terminal to all other terminals. However, this will give rise to vast cabling requirements and extensive switching systems if more than a very few terminals are being contemplated. In practice, there would probably be several transmission routes following the same pathway yet only one would be in use at any given time. An obvious financial gain would result if one or two of the transmission routes could be "shared" between all the users. However, in the event of a cable failure, this would affect communication between only two terminals in the basic system, but could affect many terminals if they share a common route. As a compromise, small groups of terminals can be connected to a switching point and then each switching point can be provided with a point-to-point link to all others to form a mesh network. This is similar to the PSTN and provides many possible routes between any two terminals. This enables the maximum use of the available cable structure, a high probability of being able to establish a communication link between any given pair of terminals and a high degree of protection from local network failures. Even with these advantages it is still far too complex a structure to use on any network much less than a national or international scale.
It is possible to connect several terminals to common transmission line such that they will each be connected to all others. Since all devices are using the same bus system it is obvious that only one communication channel can be established at any given time. It should also be apparent that some means of identifying the destination terminal must be provided unless a broadcast to all terminals is acceptable. The available bandwidth of the bus cable is shared between the user terminals and this will place a limit on the number of such terminals and of their bit rate. For this reason it is common practice to use several bus structures and interconnect them with bus extenders. Terminals that mainly communicate with each other are placed on one bus (e.g. sales department) and other groups of terminals (e.g. service department) are placed on other bus structures. Should a terminal on one bus need to communicate with a terminal on another bus, it may do so using the bus extender. In this way, much of the traffic is limited to the local bus and does not encroach on the bandwidth of other bus structures. A further advantage of using bus extenders is that the effects of noise and attenuation can be removed from the data signals as they pass between individual bus cables.
Each end of the bus must be terminated correctly to avoid reflections and all items connected to the bus must remain connected at all times, even if not in use. The choice of transmission media used for the bus will determine the attenuation and hence the maximum length.
To prevent two simultaneous transmissions taking place, one of the terminals may be designated as "master" with all others being "slave". The "master will issue permission for a slave to use the bus and will only give permission to one slave at a time. Alternatively, a system of "medium access control" (MAC) may be employed which will share out access to each terminal on an equitable basis.
In theory a bus system using bus extenders could be used to create a network of any size since the signals will be regenerated each time that they move between the local bus structures.
A modification of the single bus can be made where the two ends are brought together, the terminations removed and the resulting "free" ends jointed together. This forms a ring where all the terminals pass traffic around the ring to the next terminal. Although traffic can be transmitted in either direction, it is usual to circulate traffic in only one direction. This minimises the risk of two transmissions meeting up on the transmission route and in theory would allow several simultaneous transmissions over discrete sections of the ring. Data is passed around the ring sequentially as opposed to simultaneously in the case of a bus system. This may be relevant in time critical situations where a broadcast instruction is taking place. Because a direct link is required between each adjacent terminal, this method of connection (and the bus method) is best suited to local networks. Provision for medium access control will be required to distribute the available bandwidth between the user terminals and ensure that simultaneous transmissions do not occur and meet on the ring. Data rates of between 1 Mbps and 10 Mbps are typical for both ring and bus systems. Again, there must be some means of identifying the destination terminal (and possibly the origin) since the information will pass through each terminal on route. This information may be removed from the ring only by the destination terminal. All other terminals involved in the transmission route must simply pass the information on to the next terminal without modification.
Several terminals can be grouped on a distribution point such that each will be at the end of the transmission route. In the event of the terminal being removed, this will have no effect on transmissions between other terminals. There can be several distribution points interconnected to provide many such autonomous local star groups. The control of traffic flow will be much more complicated since each distribution point will be handling communications to and from each of its connected terminals and also traffic to and from other distribution points. Data rates will vary enormously over the various elements of the transmission network, from very low where feeding a single terminal, to very high between heavily used distribution points. If extended far enough, a star system will develop into a mesh structure similar to the PSTN or internet.
To overcome the main problem of the star system, which gave rise to very complicated switching to establish a required transmission route, the hub method uses techniques from all of the previous systems. The hub is a variation of a ring or bus such that each client terminal is connected in star configuration.
The terminals are serviced sequentially (ring) or simultaneously (bus) by the central hub. The layout looks very similar to a star and can be extended to resemble a mesh. However, because the information will be distributed to all terminals connected to the hub, there is no need to provide any switching arrangements to direct the flow to any particular terminal. The hubs themselves can become clients on larger ring or bus structures, which may then in turn be part of even larger ring or bus systems. Although there is no need for switching of signal routes, the information must carry identification as to the destination and possibly the origin.
The hub is not required to direct the information to any particular destination but simply repeats the data to all devices connected to its ring or bus. There must still be provision for medium access control to prevent simultaneous traffic from meeting at any point on the network. Sections of the system, which are connected into a bus, may be disconnected, but no part of a ring may be broken.
With a ring system there is no end to the transmission line and so no need for terminations. Clearly all terminals must remain connected as removing any one will break the flow between all others.