In Ethernet Networks, end nodes can be connected to the rest of the network primarily using two methods, namely a bus connection or a point to point connection as given in the diagram below:
Bus Connection (Broadcast channel and single collision domain)
In the bus connection method, end nodes attach to a common bus by connecting to either a Hub or a Repeater. Since a hub or repeater just repeats the signal received on one port onto all other ports, this method of connection is equivalent to connecting to a common bus. Whenever an end node sends signals, it is heard by all other nodes connected to the hub or repeater. Hence this is a broadcast channel with a common collision domain (i.e if two end nodes transmit simultaneously, then a collision results).
A sample bus connection using a hub/repeater and four end nodes is given in the figure below:
In the above figure, though four end nodes are connected to the hub or repeater in a physical topology that looks like a star, the nodes are actually connected in a bus topology (as shown in the right portion of the diagram), because the hub or the repeater broadcasts the signals received on one of its port to all other ports.
The number of nodes in a bus topology network could be increased by connecting multiple hubs in a hierarchical fashion as given in the diagram below
Though there are multiple hubs and multiple segments in the network given above, the network is still a single collision domain network, as every node hears whatever every other node transmits.
The bus connections, though popular in olden days, is not used extensively these days, due to the availability of the better point to point switched networks.
Point-to-Point Connections (Switched or Bridged)
In this method of connecting end nodes, each end node has a dedicated point to point connection to a L2 switch (also known as bridge). An example point to point switched connection with eight end nodes is given in the diagram below:
As given in the above diagram, though all the eight end nodes connect to the same L2 switch, they are still not part of a single collision domain, as the L2 switch does not broadcast signals received on one port on to all its other port. Instead a L2 switch intelligently forwards data (frames) only to the port that has the frame’s intended destination node. The L2 switch forwards data based on the destination node’s MAC address, that is present in the frame header. Due to this intelligence, each node’s link to the L2 switch is a separate collision domain. Also such point-to-point links are normally full duplex links and hence the node and the L2 switch could be simultaneously sending data to each other.
An example multiple L2 switch network
Apart from having point to point connections between end nodes and L2 Switches, there can be point to point connections between different L2 switches (normally called uplink ports), to form bigger networks with more hosts. An example of a sixteen node network interconnected by two L2 switches is given in the diagram below:
Hybrid Ethernet Topologies (Bus and point to point connections)
In order to make use of repeater or hubs that have already been purchased, some organizations and universities may still use a hybrid network consisting of both bus and point to point connections, as given in the diagram below:
In the above diagram, at the lower level, there are three separate collision domains, each served by a hub. At the next level, each hub has a point to point connection to a L2 switch.The L2 switch isolates the three collision domains and selectively forwards traffic between them. For example, if End Node 1 (EN1) sends a frame to End Node 2 (EN2), then the frame is heard by EN3 and EN4 too, as they are in the same collision domain. But the nodes in collision domain 2 and collision domain 3 do not get a copy of this frame, because the L2 switch filters out this frame based on the destination MAC address. Thus an L2 switch is able to limit the span of the collision domains in Ethernet networks and thereby improve overall throughput.