Topologies and Control Structures for Local Area Networks
Lecture 1, part 3
Network Topology
Network topology is the pattern of interconnection used among the various nodes of the network.  The most general topology is an unconstrained graph structure, with nodes connected together in an arbitrary pattern.

Advantage: arrangement of communication links can be based on the network traffic.

Cost: routing decisions have to be made at each node - substantial computation.

1) The Star Network: eliminated the need for each network node to make routing decisions by localizing all message routing in one central node. Good choice for a network with a number of secondary nodes communicating with one primary node. For example, a a number of terminals communicating with a time sharing system, in which case the central node might be the time-sharing machine itself.

If the normal pattern of communication is more general communication among all of the nodes, the central node becomes a single point of failure and possibly a performance bottleneck.

2) Ring and Bus Networks: The ring and bus topologies attempt to eliminate the central node on the network, without sacrificing the simplicity of the other nodes. Decentralized network can be constructed with a surprisingly simple structure of the nodes. In the ring topology, a message is passed from node to node along unidirectional links. There are no routing decisions to be made; the sending node simply transmits its message to the next node in the ring, and the message passes around the ring,one node at a time, until it reaches the node for which it is intended. The only routing requirement is that the node must recognize messages intended for it. Similarly, in the bus structure, there are no routing decisions required by any of the nodes. A message flows away from the originating node in both directions to the ends of the bus. The destination node reads the message as it passes by. Again, a node must be able to recognize messages intended for it.

Network Control Structures
Both the ring and the bus network introduce a problem, which is not apparent in the star net, of determining which node may transmit at any given time. In the star network, either the central node has the capacity to handle a message for every node simultaneously, or it may poll each of the other nodes in turn to determine if that node wishes to transmit. Both the ring and the bus topology must use some distributed mechanism to determine which node may use the transmission medium at any given moment.

1) Daisy Chain, Control Token, and Message Slots: the permission to use the net is passed sequentially around the ring from node to node. In daisy chain network, dedicated wires are used to pass the control information from one node to the next. Alternatively, the control information may be a special bit pattern transmitted over the regular data channel of the ring (e.g. token). A third strategy for ring control is to continually transmit around the network a series of message slots, sequences of bits sufficient to hold a full message. A slot may be empty or full, and any node, on noticing an empty slot passing by, may mark the slot as full and place a message in it.

2) Register Insertion: Another control strategy particularly suited for the ring topology is called register insertion.

 

Combination of Topology and Control Structure
Reliability Characteristics of Ring and Bus