2.7 Network topologies

---

2.7 Network topologies

As mentioned earlier, there are a variety of interconnection topologies used in local area networks. They are the global bus, the ring, and the star topologies.

2.7.1 Global bus topology

A global bus is a single shared medium, which can only be used by one device at a time. The global bus is controlled by a variety of schemes. One of the simplest is the carrier sense multiple access scheme. This protocol works by using two principles: first, the delay taken to send a bit from one end of the bus to the other and, second, the ability to send and then listen to the medium. The protocol in its simplest form operates as follows:

• Listen to the bus-if busy, wait; if clear, send data.

• Once data have been sent, continue to listen and compare what is heard against what was sent.

• If what was sent matches what is heard for the entire end-to-end communications time, then I control the bus and can continue sending a message (the assumption here is that if I wait for an end-to-end transfer time, then all other nodes must have heard my message and will now delay if they wish to transmit).

• When complete, go back into listen mode.

• If I do not hear the same message that I sent, then a collision occurred on the bus. I immediately stop transmission and delay before trying to send again.

By using this simple protocol, devices on the network can send and receive messages fairly efficiently. The problem with this protocol is that it inherently wastes media bandwidth in the sending and sensing process.

A different approach to control access to a global bus is based on a reservation scheme. In a reservation scheme the available bandwidth is broken up into chunks, which are then allocated to various devices on the network. To access the medium to transmit data a device must first wait until its reservation slot becomes available. There are numerous schemes through which the slots can be allocated and controlled. The problem with this approach is that it is inherently static. The slots cannot be reallocated easily from one system to another. Numerous variations on this protocol have been developed and implemented in systems with varying degrees of success.

2.7.2 Ring topology

The ring topology links the computer systems in the network in a continuous ring. Messages flow around the network from one computer system to another until they return to the sender. (See Figure 2.11.) This topology allows for better utilization of the medium. The medium can be broken into slots that flow around the network. The slots are marked as either empty or full depending on whether or not a message is present in the slot. To send a message a computer senses the slot beginning and checks whether it is full or empty. If the slot is full, the sender waits for the next slot. If the slot is empty, the sender inserts its message. The problem with this scheme is that the slot size limits the size of messages that can be sent in a single slot. Variations on this protocol have alleviated this problem, but have their own set of problems. A different protocol, which allows for variable-size messages, is the insertion ring protocol. This protocol requires hardware support to buffer incoming messages that would interfere with a sender's message. A computer that wants to send a message on the network can simply send the message if no other message traffic is sensed by the sender. If another message should then arrive at the sender's input during the transmission of its own message, the sender simply queues up the arriving message and appends it to the sending message when it has completed.

Figure 2.11: Ring topology.

2.7.3 Star topology

The star topology has the physical layout of a star. It has a central network processor at its center, with nodes arranged around and connected to the central point. Wiring costs can be considerably higher with this topology.

---